Road traffic is the major source of noise pollution leading to human health impacts in urban areas. This study presents the relation between changes in human brain waves due to road traffic noise exposure in heterogeneous conditions. The results are based on Electroencephalogram (EEG) data collected from 12 participants through a listening experience of traffic scenarios at 14 locations in New Delhi, India. Energetic, spectral and temporal characteristics of the noise signals are presented. The impact of noise events on spectral perturbations and changes in the relative power (RP) of EEG signals are evaluated. Traffic noise variations modulate the rate of change in α and θ EEG bands of temporal, parietal and frontal lobe of the brain. The magnitude of event-related spectral perturbation (ERSP) increases with each instantaneous increase in traffic noise, such as honking. Individual noise events impact the temporal lobe more significantly in quieter locations compared with noisy locations. Increase in loudness changes the RP of α band in frontal lobe. Increase in temporal variation due to intermittent honking increases the RP of θ bands, especially in right parietal and frontal lobe. Change in sharpness leads to variation in the RP of right parietal lobe in theta band. Whereas, inverse relation is observed between roughness and the RP of right temporal lobe in gamma band. A statistical relationship between noise indicators and EEG response is established.
Keywords: Electroencephalogram (EEG), heterogeneous traffic noise, noise exposure, psychoacoustic noise indicators
|How to cite this article:|
Manohare M, Rajasekar E, Parida M. Analysing the Change in Brain Waves due to Heterogeneous Road Traffic Noise Exposure Using Electroencephalography Measurements. Noise Health 2023;25:36-54
|How to cite this URL:|
Manohare M, Rajasekar E, Parida M. Analysing the Change in Brain Waves due to Heterogeneous Road Traffic Noise Exposure Using Electroencephalography Measurements. Noise Health [serial online] 2023 [cited 2023 Dec 3];25:36-54. Available from: https://www.noiseandhealth.org/text.asp?2023/25/116/36/372595
- Traffic noise data collection in New Delhi at 14 locations, including spectral noise measurements and audio recording.
- A listening experiment is conducted to analyse the change in brain waves due to traffic noise exposure using EEG measurements.
- The alpha and theta band fluctuation is noted in the frontal, temporal and parietal region.
- The fluctuation in brain waves is associated with an increase in loudness, temporal variation and high- and mid-frequency content of the noise signal.
| Introduction|| |
Traffic noise pollution poses several auditory and non-auditory health risks. Some of the auditory health impacts include tinnitus, noise-induced hearing loss and hyperacusis; and the non-auditory impacts include hypertension, stress, cardiovascular diseases, cognitive issues, mental load, difficulty in speech perception and understanding. Besides, the populace exposed to traffic noise pollution is more vulnerable to emotional changes. Multiple studies have reported the level of stress and annoyance caused due to noise exposure. Noise-induced stress is responsible for negative responses such as anger, anxiety, distraction and agitation. World Health Organization reports a loss of 587,000 DALYs due to noise-induced annoyance for the European Union population living in cities of more than 50,000 people. Traffic noise pollution in developing nations, including India, is one of the biggest concerns. Traffic condition in these nations is heterogeneous in nature with different operational and performance characteristics. The mix of motorized and non-motorized vehicles, different speeds and acceleration, non-adherence to lanes and high level of honking lead to a higher order of noise pollution and different kind of noise climate. This high level of noise pollution is responsible for multiple auditory and non-auditory health impacts.
Noise annoyance is one of the major health impacts due to noise pollution. Studies have demonstrated that the perceived level of annoyance is directly proportional to traffic noise pollution., However, such perception-based evaluations are highly subjective in nature. Complementary physiological measurements are essential for a comprehensive evaluation of emotional changes and health risks. An assessment of heart rate variability, respiration rate, skin conductance, eye movement, facial expression and brain wave changes are monitored in estimating the emotional changes. Functional magnetic resonance imaging (fMRI) and Electroencephalography (EEG) are employed to detect the changes in brain activity. EEG is one of the techniques which helps to analyse the real-time variation in the electrical activity of brain. EEG measurements are extensively used for stress analysis and emotion detection.
The EEG data are analysed by their frequency content. These signals are interpreted based on the power of frequency bands. The common EEG bands analysed are named delta (below 4 Hz), theta (4–7 Hz), alpha (8–13 Hz), beta (14–30 Hz) and gamma wave (31–100 Hz). The fluctuation in each of these frequency bands at specific location of the brain is associated with different activities and emotions. The change in delta band is associated with deep unconscious sleep state; theta band is associated with the change in mood, drowsiness and thought shifting; alpha band is linked with relaxed and alert mode of brain. Whereas, beta band is associated with the active, task-oriented, busy or anxious thinking and gamma band is related with working memory and concentration.
The recorded EEG signal contains multiple artefacts such as muscle movements and eye blinks, and it is required to remove this unwanted data. The preprocessing of EEG data includes filtering, sampling and artifact removal process. There is a consensus among researchers for using statistical techniques like independent component analysis (ICA), principal component analysis (PCA), multi-resolution denoising, and wavelet with higher order statistics for artifact correction. Among these methods, ICA is perceived to be a very robust method for ocular artifact removal. Discrete wavelet transform (DWT) is a time-frequency tool used to decompose the raw signal into features and for enhancing artifact suppression in EEG signals.
Studies are conducted to assess the relation between changes in power of frequency band at different lobes and changes in emotions. A lot of studies have focused on EEG asymmetry during emotion perception among individuals with different attachment patterns. For instance, the asymmetric brain activity in the frontal cortex, also called as “frontal asymmetry” is assessed to determine the changes in emotions based on valance-arousal based emotion model. The feelings associated with anger and insult are associated with higher left frontal activity and lower right frontal activity, leading to aggressive behaviours, suggested by Harmon-Jones. There is growing evidence that left frontal activation at rest and in response to emotional stimuli is associated with the approach system and positive emotion, whereas right frontal activation is related to the withdrawal system and negative emotions. Balconi and Mazza and Henriques and Davidson reported that a relative increase in left hemisphere activity was observed with positive emotional stimuli, whereas greater right hemisphere activity was associated with negative emotions.,
In the context of auditory tasks, EEG signals have been used for detecting the changes in brain activity using different types of signals, such as music and pure tones and emotional changes due to sound stimuli. A transformation in arousal rate is reported with a change in loudness level and intermittent pattern of signal. A combination of EEG features and Machine learning techniques are used to predict the emotional changes and classification of sound in arousal/valance and like/dislike categories., Thammasan et al. performed feature extraction of EEG signals using fractal dimension (FD) and power spectral density (PSD) and adopted an emotion classification algorithm to develop a music-induced emotion recognition system. The relative power (RP) of EEG signals has been employed for applications, such as automatic music recognition systems by Hadjidimitriou et al. A limited number of studies have attempted to capture the traffic noise induced changes in EEG response.
This paper evaluates the change in human brain waves caused by the exposure to traffic noise. The objectives of the study are to evaluate the neural response to traffic noise in an urban setting and identify the noise indicators which significantly represent the change in brain waves, which leads to change in emotions. The study is limited to noise signals collected from heterogeneous traffic conditions. The study analyses short-term effect of traffic noise on brain waves. The EEG data are based on listening experiment performed in a laboratory setting.
| Method|| |
The experimental procedure involved the collection of noise stimuli at 14 different locations in New Delhi, India. This included spectral and temporal noise levels recording and wave files at these locations. Descriptive noise indices were computed using the noise stimuli. The collected audio signals are used for the listening environment while recording the EEG signals. The workflow of the study is explained in [Figure 1].
|Figure 1 Flowchart explaining experimental and analysis process of the paper.|
Click here to view
Traffic noise data collection
In all, 14 different locations in New Delhi, India were selected for data acquisition (indicated as P1 - P14), shown in [Figure 2]. These locations could be classified into two zones, Zone 1 (Shahdara) and Zone 2 (Karol Bagh). Zone 1 is a residential area adjacent to a major arterial road and Zone 2 is one of the major commercial business districts of the city. These locations were representative of different traffic conditions found in a heterogeneous transportation system. Selected locations vary from the arterial, sub-arterial, collector and local streets, with varying composition and density of road traffic. The data are collected in the month of January 2020. Initially, peak hours at selected locations are identified through available noise data from Central Pollution Control Board, India (CPCB). Accordingly, the measurements are carried out during day time in three segments: P1–P5 (9–11 am), P6–P10 (12.30–2.30 pm) and P11–P14 (4–6 pm) [Table 1].
|Figure 2 (a) shows different locations adopted to record the noise levels in New Delhi. (b) Showing the whole setup as recording the noise levels at one of the selected locations. (c) Richo Theta 360° camera. (d) 3-DIO binaural recorder used to record an audio file. (e) Delta Ohm Class 1 sound level meter.|
Click here to view
|Table 1 Description of each location with land-use, street type and its picture|
Click here to view
Noise levels were measured using a class I sound spectrum analyser (Delta Ohm) at 0.5 second sampling interval along with spectral content in a 1/3 octave band in the range of 16 Hz to 8 kHz. The duration of recorded stimuli was 300 seconds at every location. Simultaneously, 3Dio Binaural Audio recorder was used to record noise signals in (.wav) file format. The Binaural recorder was mounted at the height of 1.5 m and placed clear of any reflecting vertical surfaces. Along with this setup, the omnidirectional camera is used to record the events happening during the recording period. [Figure 2] shows the setup of instruments used to record noise levels.
Environmental noise descriptors considered for this study are A-weighted equivalent sound pressure levels (LAeq), percentile noise levels (L10, L90), frequency content of the signal (16 Hz to 16 kHz) and Sharpness, Loudness (ISO 532-1), temporal variability of the noise signal (L10–L90 and fluctuation strength). LAeq was computed for 300 seconds time period with sound pressure levels sampled at 1 second intervals. L10 and L90 reflects the temporal variation (fluctuation) of the noise signal and is also defined as Noise Climate (NC). The whole frequency spectrum was divided into low-, mid- and high-frequency ranges. Three indices LAeq (LF), LAeq (MF) and LAeq (HF) were computed. LAeq (LF) represents the noise equivalent levels for the frequency band 16 to 250 Hz, LAeq (MF) represents the noise equivalent levels of mid-frequency band ranging from 250 Hz to 2 kHz. The LAeq (HF) represents the noise equivalent values of high-frequency band ranging from 2 to 16 kHz. The recorded binaural wave files were used for the calculation of psychoacoustic parameters such as Loudness, Fluctuation strength, Roughness and Sharpness using MATLAB Audio toolbox.
In all, 12 participants (8 males, 4 females aged between 22 and 28 years old; mean age = 25 years; and standard deviation = 2.60) participated in the experiments. The number of participants has been considered in accordance with the sample size of previously published studies, and power analysis. All participants were university students pursuing post graduate degree and reported normal hearing with no known ailments. The experiment was conducted at the Indian Institute of Technology, Roorkee.
EEG data collection
The listening experiment was performed in a laboratory setting with a background noise not exceeding 35 dB(A) and RT60 1.1 s. Each participant was individually subjected to a listening experience of the traffic situation and the EEG signals were recorded using a 14-channel recorder (Emotiv EPOC+). The binaural audio experience was provided using active noise cancellation earphones. Multiple studies have reported the reliability of auditory ERPs using Emotiv EPOC+ EEG system., The listening experiment is developed based on the stereophonic sound reproduction technique in the laboratory., This procedure is widely acknowledged for its good representation, readability, plausibility and overall reproduction quality for fixed and moving noise sources.,
[Figure 3] shows the arrangement of different devices used in the experiment. The binaural audio was presented to the participant through the high-definition noise cancellation earphones (Sony WI-1000×). All 14 traffic noise stimuli were presented to each participant in a randomised order. The length of each noise stimulus was 5 minutes and EEG data were recorded for the same period. For analysis, the first and last minute of the recording is discarded and reaming 3 minutes recorded data are used. Before the start of the experiment, each participant was provided with a dummy experiment, where the handiness of the participant was checked. After confirming the experimental conditions as comfortable, the experiment was initiated. At the start of experiment, the baseline data of the participant were recorded in two conditions, “eyes open” and “eyes closed” for 30 seconds each. These baseline data were used to normalize the EEG response of all participants while performing the analysis. While listening to the noise stimuli, participants were instructed to focus on a green dot on a blank screen with minimal-to-no body movement. After listening to each signal, the participant was asked to rate their perception in the terms of annoyance caused. The EEG headset recorded the signals generated due to changes in the electrical activity of the brain. The magnitude of these electrical activities is minimal, ranging from 5 to 100 μV. The headset consisted of 14 electrodes placed over the scalp of human brain according to 10 to 20 international system. The 14 electrodes in EEG headset are referred to as AF3, F7, F3, FC5, T7, P7, O1, O2, P8, T8, FC6, F4, F8 and AF4 along with two reference electrodes, as shown in [Figure 3](C). The nomenclature of electrodes depends upon the location of electrodes. The later denotes different lobes of the human brain, that is, F − frontal lobe, T − temporal lobe and O − occipital lobe. The number denotes the position of electrodes − an even number denotes the right side and an odd number denotes the left side of the human brain.
|Figure 3 (a) The schematic setup of different equipment used in a listening experiment. (b) The actual photograph showing experimental setup in a laboratory. (c) Location of different electrodes in 10–20 international system positions.|
Click here to view
A schematic overview of the experiment is presented in [Figure 4]. The experiment was followed by a paper-based survey to assess the overall experiment and to know any difficulty faced during the experiment. The whole experiment lasted for an average of 115 minutes per participant. Participants were allowed to interrupt and leave the experiment in case of any difficulty faced while experiencing the signals.
|Figure 4 Flowchart explaining the process and steps undertaken in a listening experiment.|
Click here to view
EEG data processing
The raw EEG signals include multiple artefacts like muscular activities, eye blinks and noise due to electrical interference. These noises and artefacts affect the useful features in the original signals. Hence, it is necessary to remove the unwanted components in the raw EEG signal. The magnitude of original signals ranged between 5 and 100 μV, which is smaller than the magnitude of artefacts. To remove the noise from this signal, the following steps are adopted:
The finite impulse response (FIR) digital filter was used for band-pass filtering. This removes the out-of-band signals and considers only essential signals. The fifth order FIR filter was used with a high-pass filter of 50 Hz and low-pass filter of 0.5 Hz at the sampling rate of 128 Hz. Zero-phase digital filtering was achieved by using the FIR filter in a forward and backward direction. A data smoothening was performed using a moving average (MA) filter at fifth order. MA filter is typically used to remove the unwanted peaks and behaviour of the signal. In this process, every point in the signal is replaced by the average of the defined set of point. Pre-processing of the raw signal was performed using MATLAB.
The Equation (1) for moving average filter is as follow:
where S is the input signal, y is the output signal, A is the order of the moving average.
Independent component analysis
It is a statistical technique which represents a multidimensional random vector as a linear combination of non-Gaussian random variables (“independent components”) that are as independent as possible. ICA is a non-Gaussian version of factor analysis and somewhat similar to principal component analysis. The statistical independence of the estimated component increases with ICA. Raw EEG signal is a combination of different artefacts and noise, and the sources are combined linearly in an electrical field at the electrode level. Hence, ICA is required for EEG signal. In this study FastICA technique was implemented, using MATLAB.
The component “xi” of the dataset x = (x1,…, xm)T is created as an addition of the independent components “sk,” where k = 1,…, n by using the Equation (2).
where “ai, k” is mixing weights. Considering “W” as the transformation weight matrix, the ICA can be calculated using Equation (3).
After performing ICA, the signal was processed with a discrete wavelet transform for feature extraction. As the EEG signal is a time-domain signal, its characteristics are embedded in the noise and need to be converted from time-domain to frequency-domain to derive useful features from it.
Discrete wavelet transform (DWT)
DWT was used to convert a signal into smaller waves through multi-stage decomposition. The fundamental principle behind the wavelet analysis is to decompose the signal S into a number of wavelet coefficients, after which the signal can be interpreted as a linear combination of the wavelet weighted by the wavelet itself. In DWT, the signal is processed at multiple frequency bands, with the different resolution by decomposing the signal into approximations and detail coefficients. These approximations are further decomposed with a similar decomposition process, by applying high-pass and low-pass filter to the time-domain signal.
The high-pass and low-pass filter was applied to signal using the following equation:
The procedure for DWT for the EEG signal S is shown in [Figure 5]. A decomposition signal at eight levels was conducted using the orthonormal wavelet technique Daubechies-8 (DB8). Each level consists of two digital filters (HP − high-pass filter and LP − low-pass filter) and two down-samplers, which compress the signal to half of its size. The output of every level is a detailed coefficient (Di) and approximation coefficients (Ai). The approximation coefficient received at a previous level is further used for the same decomposition process at the next level, till the required level of decomposition is achieved.
|Figure 5 Decomposition of EEG signal using orthonormal wavelet Daubechies-8 (DB8) technique.|
Click here to view
The energy of detailed coefficient (Di) and approximation coefficients (Ai) was calculated as follows:
where l is decomposition level, N is the number of details and approximation coefficients at each level. EDi and EAi are the energy level of the detail coefficient and approximation coefficient level l.
The DB8 decomposition provides the first four wavelet coefficients that represent noise and five wavelet coefficients that represent frequency bands defined as, gamma − (>30 Hz), beta (14–30 Hz), alpha (7–13 Hz), theta (4–7 Hz), delta (<4 Hz). [Table 2] presents the activities related to the four major frequencies of the brains. Generally, these activities were shown to interact with specific brain frequencies.
|Table 2 Association of different activities with the different frequency band of brain waves|
Click here to view
After DWT, the power and RP of each electrode for every subject were calculated for DWT coefficients. Such features were previously used in EEG-based studies and have provided reliable results., Power and RP can be analysed in the time-frequency domain and calculated using Equations (8) and (9), respectively. Band power (x) returns the average power in the input signal, x. If x is a matrix, then band power computes the average power in each column independently.
Power of signal x
RP of signal x
The outputs of electrodes in each brain lobe were averaged together to get the RP of that particular lobe. The mean of all seven electrodes present on the right side was considered for the RP of the right lobe. Similarly, the mean RP of electrodes present on the left hemisphere was considered for the RP of the left lobe. In addition, the overall RP was used for the analysis, which is the mean RP of all 14 electrodes. As every EEG signal was decomposed into four bands, the RP was also calculated for each separate frequency band. [Table 3] shows the different electrodes used for averaging the RP of each brain segment.
|Table 3 The list of electrodes used for determining the RP of each brain lobe|
Click here to view
| Results|| |
Characterisation of traffic noise stimuli
[Figure 6] shows the variation of noise levels and the spectral content for the noise signal at P1 and P14. The stimuli P1 represent a quieter location and P14 represent a noisier location. Location P1 experienced relatively less traffic volume with a nearly continuous flow of traffic without any honking. Location P14 experienced a high volume of traffic, and heterogeneity in vehicular type and traffic flow.
|Figure 6 Comparison of two extreme stimuli P1 and P14 based on temporal and spectral indicators.|
Click here to view
[Table 4] displays the energy content, temporal variation and spectral content of the signal for all 14 locations that have been analysed. The relation among noise indicators, evaluated through Pearson correlation test, is presented in Appendix A.
|Table 4 Summary of energy, temporal and spectral content of traffic noise signal at 14 locations|
Click here to view
A significant difference in noise stimuli across locations is confirmed through a one-way ANOVA test performed on the LAeq, yielding an F value (3.40) higher than F critical (1.78) with P < 0.005. The highest LAeq was recorded at P14 (80.98 dB) and the lowest LAeq was recorded at P1 (70.15 dB). The minimum noise level of 58.80 dBA and the maximum noise level of 101.20 dBA is recorded throughout the locations. Maximum loudness (N) was observed at P14 (65.71 sones) and the minimum loudness was recorded at P1 (51.16 sones). As both the parameters are dependent on each other, high collinearity is observed in both the indicators (r = 0.81, P < 0.01).
The temporal variation (L10–L90), fluctuation strength and roughness are significantly related with change in LAeq of signal; (r = 0.81, P < 0.01 for FLS and r = −0.64, P < 0.05 for R). The L10–L90 was maximum at P14 and minimum at P1. The change in LAeq and L10–L90 is moderately correlated (r = 0.54, P < 0.05) and no significant correlation is observed between the change in loudness and L10–L90. FLS increases with an increase in LAeq of the signal, whereas R decreases with an increase in LAeq of the signal. The FLS is significantly correlated with mid-frequency (r = 0.66, P < 0.01) and high-frequency (r = 0.76, P < 0.01) content of the signal. Whereas, R is significantly correlated (r = −0.70, P < 0.01) with mid-frequency content only.
The subjective survey is conducted to assess the level of annoyance perceived by the listener during the experiment. Participants were asked to report the level of annoyance caused on the scale of 0–10, where “0” is considered as “not annoyed” and “10” is considered as “highly annoyed.” Pearson Correlation test is conducted to identify the relation of perceived annoyance and noise indicators. A statistically significant correlation between LAeq and annoyance is observed with r = 0.69, P < 0.01. Also, strong positive correlation is observed between LAeq (MF) and annoyance, with r = 0.67, P < 0.01. From these results, it is inferred that with an increase in LAeq and mid-frequency content of the signal, annoyance level also increases. Thus, the stimuli with lowest LAeq (P1) are considered as a least annoyed signal and stimuli with highest LAeq (P14) are considered as a highly annoyed traffic noise signal.
Changes in EEG signals due to noise exposure
The impact of noise on EEG signals is presented in two sections. The first section examines the changes in the power of EEG signals of participants, considering two distinct stimuli, P1 and P14. It includes changes in power spectral density of EEG data across all electrodes followed by a comparative analysis using the event-related spectral perturbation (ERSP) approach. The second section evaluates the change in the RP. The RP of each electrode is averaged across all participants and classified into four frequency bands (i.e., θ, α, β and γ). The change in the RP is examined with respect to electrode location and different brain segments. This analysis is carried out across all 14 noise stimuli.
Power spectral density
The PSD is analysed for change in the power of EEG signals for two distinct scenarios with short-term traffic noise exposure. The channel spectrum and scalp topographical maps for different frequency bands are obtained using the EEGLAB toolbox. In EEG lab Toolbox, the raw data are filtered with FIR filter within the range of 0.5 to 50 Hz and moving average smoothening is applied at the fifth order. The data are referenced with the average function, and independent component analysis is performed using “runica” function. [Figure 7] presents the topographical maps of EEG response of a subject corresponding to stimuli P1 and P14. Each coloured trace represents the spectrum of the activity of one data channel. The power spectral density is plotted at different frequencies. The scalp topography is analysed at 6, 10, 12, 22 and 32 Hz to replicate θ, α, β and γ frequency bands, respectively. P14 induced higher power levels compared with P1 for a similar frequency range. The electrodes in the frontal and right parietal lobe yielded more activate responses as compared with other electrodes.
|Figure 7 Power spectral density and associated topographical maps for scenario P1 and P14.|
Click here to view
Event-related spectral perturbation
Event-related spectral perturbation depicts the spectral power of different electrodes over a time period in a broad frequency range. It represents the mean event-related changes in the form of the spectrogram of different electrodes. Calculating ERSP requires computing the power spectrum over a sliding latency window then averaging across data trials. The colour at each image pixel then indicates power (in dB) at a given frequency and latency relative to the time locking event. Typically, for n trials, if Fk(f, t) is the spectral estimate of trial k at frequency f and time t.
To compute Fk(f, t), EEGLAB uses either the short-time Fourier transform, a sinusoidal wavelet (short-time DFT) transform or a Slepian multi-taper decomposition that provides a specified time and frequency resolution.
[Figure 8] presents the ERSP of different electrodes (AF3, AF4, O1, O2, T7 and T8) for two stimuli P1 and P14. The ERSP is plotted as a spectrogram for the frequency ranging from 0.5 to 50 Hz. The temporal variation of traffic noise levels is placed below the respective plots. The major events impacting the ERSP are marked in red.
|Figure 8 The ERSP plot for stimuli P1 and P14. The majorly affecting electrodes F3, F4, T7, T8, O1 and O2 are used for comparison along with waveform of P1 and P14 stimuli.|
Click here to view
The magnitude of ERSP increased with each instantaneous increase in traffic noise. The honking events excite the signals that increase the power for a short period. The electrodes in the parietal lobe followed a similar pattern as that of the frontal lobe, where the overall power increased with an increase in noise levels. The temporal lobe, which is responsible for auditory processing, was analysed using electrodes T7 and T8. Due to the relatively low noise level at P1, the magnitude of ERSP was lower, whereas it increased at P14. The impact of noise peaks was clearly observed in P1, whereas for the location P14, the distinct impact of noise peaks was not clearly observed due to an increase in overall noise levels. Similarly, the occipital lobe was analysed with O1 and O2 electrodes. The impact on the occipital lobe was more in P14 due to the large number of vehicles and honking, which were easily noticed by the listener. Studies have explained the activation of the visual cortex for sudden and noticeable noise events in an auditory task.
RP variation in EEG sub-bands
The change in RP was averaged across all participants for stimulus P1 and P14. The RP at γ, β, α and θ frequencies was considered for the comparison. RP(θ) for location P1 was considerably smaller than P14 for all electrodes. RP(α) at P1 was lower for all electrodes except for electrodes AF3, AF4, F7 and F8, indicating the activation of the alpha band in the frontal lobe of participants. The electrodes of temporal, occipital and parietal lobes had a higher range of RP(α). The RP(β) for all electrodes was higher for P14. The RP(θ) was consistently higher in all electrodes with an increase in noise levels. In contrast, RP(β) increased for electrodes of the temporal, occipital and parietal lobes. [Table 5] presents a summary of the RP (γ, β, α and θ) variations of the four lobes for stimulus P1 and P14. [Figure 9] compares the RP for stimuli P1 and P14 at different frequency ranges using 2-D brain topography plot.
|Table 5 Relative power averaged across all participants for P1 and P14 noise stimuli, and categorised into different frequency bands|
Click here to view
|Figure 9 Scalp topological distributions for P1 and P14 noise stimuli showing change in magnitude of relative power of EEG signal, for frequency bands delta (1–4 Hz), theta (5–7 Hz), alpha (8–12 Hz), beta (13–29 Hz) and gamma (30–50 Hz).|
Click here to view
RP variation across locations
[Figure 10](a) presents the RP (γ, β, α and θ) variation for the frontal, temporal, occipital and parietal lobe. [Figure 10](b) presents the RP (γ, β, α and θ) variation for the right and left hemispheres of the brain. RP(θ) and RP(α) increase with an increase in the overall loudness of the noise signal.
|Figure 10 Figure A. showing normalized values of relative power for frontal, partial, Temporal and Occipital lobe., Figure B mean relative power and relative power of Left and Right Hemisphere.|
Click here to view
RP of the frontal lobe was lowest for all frequency bands, followed by the temporal lobe. The RP of EEG signals in the occipital and parietal lobe was higher across frequency bands followed by temporal lobe and lowest in frontal lobe. The rate of variation of θ was less compared with α and β. Maximum variation of α and β was observed in P04 and P12. The amplitude of RP was lower in the left lobe compared with the right lobe and overall RP. The maximum deviation in RP was observed in α and β bands, whereas the θ band had a relatively low rate of RP fluctuation.
Relation between noise indicators and RP of EEG signals
The relation between different noise stimuli and change in RP at different lobes for four frequency bands is analysed through Pearson correlation. [Table 6] shows the correlation (r) between the noise indicators and RP at different lobes.
|Table 6 Relation between Noise indicators and relative power at different brain regions|
Click here to view
An increase in the overall loudness of traffic noise and the temporal variation in high-frequency range of the noise signal affected the θ band. This is evident from the correlation exhibited by RP(θ) with most of the studied noise indicators. The overall RP(θ) have a significant relation with L10–L90, LAeq (MF), and LAeq (HF) with r = 5.55, 0.55 and 0.58 (P < 0.05), respectively.
The RP of the left hemisphere at theta band, that is, RP(LH,θ) exhibits a significant relationship with loudness and LAeq (MF) with r = 0.61 and 0.65 (P < 0.05), respectively. Similarly, the RP of the right hemisphere at theta band, that is, RP(RH,θ) shows significant relation with L10–L90, LAeq (HF), and sharpness with r = 0.62, 0.61 and 0.57, respectively. The relation between right and left frontal lobe at the theta band is observed with noise indicators. A significant relationship is observed between RP(F,θ)L and Loudness and LAeq (MF) with r = 0.59 (P < 0.05) and r = 0.67 (P < 0.01), respectively. RP(F,θ)R has a significant relationship with L10–L90, LAeq (HF), and sharpness with r = 0.56, 0.65 and 0.59 (P < 0.05), respectively. RP at left and right parietal lobe has also presented a significant relationship, where RP(P,θ)L has a significant relationship with loudness (r = 0.65, P < 0.05) and LAeq (MF) (r = 0.65, P < 0.05). Also, RP(P,θ)R has a strong positive relation with L10–L90 (r = 0.70, P < 0.01) and sharpness (r = 0.64, P < 0.05). LAeq and LAeq (HF) also have a positive relation with RP(T,θ)R, where r = 0.55, 0.66, P < 0.05, respectively. Whereas, roughness has shown a strong negative relation with RP(T,θ)R with r = (−0.56), P < 0.05.
The alpha band has presented strong positive relation with loudness and temporal variation of traffic noise. The RP(F,α), RP(F,α)L, RP(P,α)R, RP(T,α)R have a positive association with loudness (N) with r = 0.56, 0.58, 0.56, 0.56 (P < 0.05), respectively. The indicator L10–L90 has depicted the association with RP of the right temporal at alpha band with r = 0.64, P < 0.05. Apart from loudness and temporal variation, no significant association is noted with alpha band.
Likewise, RP(β) has a statistically significant association with temporal variation and sharpness of the traffic noise signal. Where RP(RH,β), RP(F,β)R, and RP(T,β)R are positively correlated with L10–L90 with r = 0.56, 0.56 and 0.60 (P < 0.05), respectively. Also, the RP of right occipital lobe at beta band is correlated with sharpness with r = 0.61, P < 0.05.
The RP of gamma band is negatively associated with roughness and positively associated with the mid-frequency content of the signal. From the analysis, it is observed that RP(RH,γ), RP(P,γ)R, and RP(T,γ)R have a strong negative correlation with roughness with r = (−0.56), (−0.63) and (−0.66), P < 0.05, respectively. In addition, the RP(P,γ)R and RP(T,γ)R have a positive correlation with LAeq(MF) with r = 0.56 and 0.53, P < 0.05, respectively.
Stepwise linear regression model
The relation between different noise indicators and the RP of EEG signals was analysed through a stepwise regression. The RP of each lobe was averaged across 12 participants for each location. Table 7 presents the summary of stepwise regression models established between noise indicators and the RP of EEG signals at different locations.
LAeq, N, L10–L90, LAEQ (MF), LAeq (HF), sharpness, and roughness explain significant variation in different parts of the brain. The theta and gamma frequency bands have established a significant relationship with noise indicators. LAeq significantly impacted θ band in the right temporal lobe (r = 0.55, t = 2.26, P < 0.05), explaining 30% variance in RP (T,θ)R. A unit increase in LAeq has resulted in 0.55 unit increase in RP (T,θ)R. Temporal variation of the noise signal impacts the change in RP at the tight parietal lobe. It is analysed through the relationship between L10 and L90 with change in theta band of right parietal lobe with L10–L90 explaining 48% variance of RP (P,θ)R (r = 0.70, t = 3.35, P < 0.05). A unit change in L10–L90 results in 0.7 unit variation of RP (P,θ)R.
The mid- and high-frequency bands have shown a significant relationship with change in the RP at frontal lobe. The variation in mid-frequency band seems to have impacted the left frontal lobe in theta band. The variation in high-frequency band has impacted the variation in RP at right frontal lobe in theta band. LAeq (MF) has a significant relation with RP (F,θ)L (r = 0.67, t = 3.10, P < 0.01), explaining 45% of variance. A unit change in LAeq (MF) reflects 0.67 units change in RP (F,θ)L. Similarly, LAeq (HF) shows significant relation with RP (F,θ)R (r = 0.65, t = 2.95, P < 0.05). Change in one unit of LAeq (HF) results in 0.65 units of change in LAeq (HF).
The psychoacoustic indicators have exhibited significant relation with change in the parietal and temporal lobe in theta and gamma band. The change in loudness and RP at the left parietal lobe has a significant relationship with (r = 0.64, t = 2.54, P < 0.05), with 42% of variance explained. A unit change in the level of loudness reflects in 0.64 unit change in RP (P,θ)L. Similarly, the relation between sharpness and RP at right parietal lobe has a statistically significant relation with (r = 0.64, t = 2.91, P < 0.05), explaining 41% of variance. From this model it is noted that with a unit change in sharpness, there will be 0.64 unit change in RP(P,θ)R.
| Discussion|| |
In this study, we examined the variation in brain waves due to traffic noise exposure. The following points can synthesize this study. The noise signals were dominated by low-frequency (engine noise) and high-frequency noise (honking noise). The highest value of sharpness (S) was observed for the scenarios that are dominated by high-frequency content, that is, honking. Increased honking leads to an increase in sharpness and L10–L90 of the signal. The energy content, the rate of amplitude variation and the spectral content of the signal play a vital role in explaining the character of the noise signal. The degree of intermittency in the honking impacts the change in the temporal variation of the noise signal. The overall loudness, as well as the temporal variation (i.e., change in peaks and drops over time) are essential for a realistic representation of the noise scenario.
The recorded EEG signals were analysed to establish a relation with noise indicators. Firstly, the change in power spectral density of the signal was analysed for two distinguish traffic noise stimuli and its effect on the listener was determined. This is followed by assessing the relationship between the EEG frequency band at different lobes and noise indicators through correlation and regression analysis.
As analysing the change in brain response of two distinct stimuli, it is observed that with an increase in traffic noise level, the power in θ and α of left frontal lobe electrodes reduces and the power in θ and α of right parietal lobe electrodes increases. The fluctuation in power of electrodes in the frontal lobes corresponds to a drowsy, bored and annoyed state of the subject., The β band of the frontal lobe, right parietal lobe and right temporal lobe electrodes was activated for location P1, compared with location P14. A similar trend was observed for γ band. The increase in noise levels yields a higher PSD in the temporal lobe. This can also be attributed to the higher engagement of the temporal lobe while listening to traffic events. The increase in EEG signal power of right hemispherical region of the brain indicated the change in emotion toward the negative side, as suggested by Ahernt et al. EEG signal power increased for the noise peaks which occurred after a long silence. Whereas, the EEG signal power stabilised when the traffic noise was at a continuously higher level. The scenario can be interpreted in a manner that presumed noise events have less impact on the human brain compared to discrete noise events.
Considering the contextual factors at different locations, the changes in brain waves are analysed. It is observed that for theta, alpha and beta bands, the fluctuation of RP in right hemisphere is more as compared with the left hemisphere. This high level of fluctuation in the right hemisphere of the brain indicated the shifting of the mental workload of the participants. In multiple studies, it is concluded that the left hemisphere is associated with the positive emotion stimuli and right hemisphere activities are associated with negative emotions.,
From the Correlation analysis, the variation in the theta band is commonly observed over all lobes of the brain. The temporal variation and high-frequency content which are contributed due to traffic have an impact on change RP(θ) at the frontal, parietal and temporal lobes of the brain. Also, the chattering of people and hawkers, which contributes to the overall high-frequency content of soundscape, leads to the change in theta band over the frontal, temporal and parietal lobes. This increased theta power is associated with the decreased reaction of the cortex to the sensory stimuli, which is observed in this experiment. In one of the studies conducted by Teplan et al., it is stated that theta band deals with integrative cognition and association functions.
Change in alpha band is observed over the frontal, temporal and parietal lobes. The change in frontal alpha band leads to a change in emotion, as previously demonstrated by various studies. The greater left frontal lobe is mainly associated with anger increase in working memory load and this fluctuation is highly correlated with the traffic noise exposure of the subject.
L10–L90 of the signal which is due to intermittent honking have impacted beta band at the right hemisphere of the brain, specifically the frontal and temporal lobe. Change in sharpness, which is associated with high-frequency content of the signal, has impacted the right occipital lobe. The beta band is associated with stress and has a strong association with a temporal variation of signal. This relation is associated with the experience of a stressful environment of the subject with negative arousal.
The change in RP of gamma band is observed due to fluctuation of roughness in the signal. This association is negatively correlated, which indicates that with an increase in roughness, the RP of gamma band at the right temporal and parietal lobe decreases. The gamma band is essential for cognition, perception and memory processing. The roughness is often considered as a negative indicator of the soundscape quality, that is, higher roughness leads to poor and annoying soundscape perception. Thus, this inverse relation between the fluctuation of gamma band and roughness indicates that roughness content of traffic noise impacts listeners negatively.
In the stepwise regression modelling, different independent models are established between noise indicators and change in RP. It is found that the overall noise levels have an impact on the temporal lobe of the brain, which is directly associated with the auditory processing. The high level of temporal variation is often considered as an indicator leading to a chaotic noise environment. The association of temporal–parietal junction with the infrequent noise stimuli was also studied by Knight et al. This indicates that intermittent honking and presumed noise events have an impact on the temporal and parietal lobes, leading to a negative perception of traffic noise environment. Fluctuation of theta band in the frontal lobe is significantly associated with mid- and high-frequency content of the traffic noise signal, respectively. This fluctuation in frontal theta is associated with increased cognitive load and fatigue., From these results, it can be concluded that the high level of mid- and high frequency emitted by road traffic cause fatigue and increased cognitive load.
Multiple studies have analysed the relation between sound waves and the activation of brain waves in different regions, which can help to derive the subject’s emotional state. It is observed that while listening to the music, the cortical activation is noted, increasing the participants’ arousal state. The frontal lobe is mainly associated with an increase in working memory load, and this fluctuation is highly correlated with the traffic noise exposure of the subject. The phenomenon of enhancing the right hemisphere activation is reported in the experiment, which is linked to altered states of consciousness. The frontal cortex also has a response-controlling function. The increase of alpha rhythms, which is impacted due to an increase in loudness, is functionally correlated to several types of cognitive, sensory and motor behaviours. Few studies have suggested that the gamma-band response is related to selective attention. The significant negative relation with roughness and gamma-band can explain the change in R, which explains the change in the subject’s attention while listening to the noise signals. The current findings are in agreement with Jafari et al., where a similar change in RP is observed with a change in the amplitude of the noise signal. Also, an increase in subjective annoyance was noted for high-frequency noise and an increase in energy of beta and theta components is noted for unpleasant music by Li et al. and Balasubramanian et al.,
| Conclusion|| |
The paper presents the relationship between fluctuations of brain waves due to exposure to heterogeneous traffic noise signals. The results are based on EEG data collected through the listening experience of 12 participants. Characterisation of the measured noise data is presented in terms of energetic, spectral and temporal attributes. The changes in power spectral density of EEG data followed by a comparative analysis using the event-related spectral perturbation (ERSP) approach for select locations were presented. Change in RP of θ, α, β and γ frequency bands with respect to different brain segments was examined. The relation between noise indicators and the RP of EEG signals was presented.
Considering the finding, it can be concluded that overall sound pressure level, temporal variation and the frequency content of the traffic noise strongly influence the change in brain waves, which can lead to changes in emotions. Intermittent and intense honking pattern with high-frequency noise content is responsible for major changes in the brain waves of the listener. The change in temporal variation and the spectral content of the signal is responsible for the inactive and drowsy state of the subject while hearing the traffic noise signals. This is evident from the strong association of L10–L90, LAeq (HF) and LAeq (MF) with the theta band activities. The change in RP(F, α), which indicates a change in a comfortable, quiet and coherent state of mind, is strongly associated with loudness of signal. Moderate-to-strong correlation of different noise indicators with α and θ bands highlights the impact of traffic noise exposure on change in comfortable and quiet state of the listener, leading to drowsiness and annoyance. As the high frequency content and loudness are highly correlating noise indicators with RP(θ) and RP(α), it may be inferred that honking and engine acceleration noise are responsible for an increase in annoyance and restlessness of the participants. This also leads to changes in concentration, relaxed mood and increase in anxiety of participants.
Further studies are required to understand the relation between level of intermittency in traffic noise and the human brain response. The study can be extended to analyse the long-term effect of noise exposure and the effect of noise sensitivity on EEG signals. Similar experiments can be performed to analyse the effect of audio visual scenarios, effect of different socio-demographic groups to find the impact of soundscape stimuli on listeners.
The author is grateful to all the participants for participating in the experiment. Author would like to thank Mr Aditya Saini and Mr Subramanian G for help during data collection and Mr Bhavya Garg for the valuable inputs during data processing.
Financial support and sponsorship
The research work is supported under Prime Minister Research Fellowship, Ministry of Education, India (PMRF ID: 2800113).
Conflicts of interest
The authors state no conflict of interest.
| References|| |
WHO. WHO: Burden of disease from environmental noise. World Health 2011;1-105.
Murugappan Z, Yaacob M. “Methods and approaches on inferring human emotional stress changes through physiological signals: a review. ” Int J Med Eng Inform 2013;5:152-162.
Annerstedt M, Jönsson P, Wallergård M et al.
Inducing physiological stress recovery with sounds of nature in a virtual reality forest — results from a pilot study. Physiol Behav 2013;118:240-50.
Yu B, Hu J, Funk M, Liang R, Xue M, Feijs L. RESonance: lightweight, room-scale audio-visual biofeedback for immersive relaxation training. IEEE Access 2018;6:38336-47.
Hedblom M, Gunnarsson B, Schaefer M, Knez I, Thorsson P, Lundström JN. Sounds of nature in the city: no evidence of bird song improving stress recovery. Int J Environ Res Public Health 2019;16:1390. https://doi.org/10.3390/ijerph16081390
Gonzalez-Sanchez J, Chavez-Echeagaray ME, Atkinson R, Burleson W. ABE: an agent-based software architecture for a multimodal emotion recognition framework. Proceedings − 9th Working IEEE/IFIP Conference on Software Architecture, WICSA 2011 2011: 187–93. https://doi.org/10.1109/WICSA.2011.32
Khushaba RN, Greenacre L, Kodagoda S, Louviere J, Burke S, Dissanayake G. Choice modeling and the brain: a study on the Electroencephalogram (EEG) of preferences. Expert Syst Appl 2012; 39: 12378–88.
Yuan Q, Liu XH, Li DC, Wang HL, Liu YS. [Effects of noise and music on EEG power spectrum]. Space Med Med Eng 2000;13:401-4.
Blanco JA, Vanleer AC, Calibo TK, Firebaugh SL. Single-trial cognitive stress classification using portable wireless electroencephalography. Sensors 2019;19:1-16.
Woaswi W, Hamzah N, Omar W, Saifuddin A, #1 WI, Hanif M. Human emotion detection via brain waves study by using electroencephalogram (EEG). Int J Adv Sci Eng Inform Technol 2018;6:1005.
Jung TP, Makeig S, Westerfield M, Townsend J, Courchesne E, Sejnowski TJ. Analysis and visualization of single-trial event-related potentials. Hum Brain Mapp 2001;14:166-85.
Lagerlund TD, Sharbrough FW, Busacker NE. Spatial filtering of multichannel electroencephalographic recordings through principal component analysis by singular value decomposition. J Clin Neurophysiol 1997;14:73-82.
Kumar PS, Arumuganathan R, Sivakumar K, Vimal C. Removal of ocular artifacts in the EEG through wavelet transform without using an EOG reference channel. Int J Open Problems Compt Math 2008;1.
Jayaraman S, Krishnaveni V, Jayaraman S, Aravind S, Hariharasudhan V, Ramadoss K. Automatic identification and removal of ocular artifacts from EEG using wavelet transform. Meas Sci Rev 2006;6.
Ghandeharion H, Erfanian A. A fully automatic ocular artifact suppression from EEG data using higher order statistics: improved performance by wavelet analysis. Med Eng Phys. 2010;32:720-9.
Castellanos NP, Makarov VA. Recovering EEG brain signals: artifact suppression with wavelet enhanced independent component analysis. J Neurosci Methods 2006;158:300-12.
Harmon-Jones E. Contributions from research on anger and cognitive dissonance to understanding the motivational functions of asymmetrical frontal brain activity. Biol Psychol 2004;67:51-76.
Gander M, Buchheim A. Attachment classification, psychophysiology and frontal EEG asymmetry across the lifespan: a review. Front Hum Neurosci 2015;9:79.
Henriques JB, Davidson RJ. Decreased responsiveness to reward in depression. Cogn Emot 2000;14:711-24.
Balconi M, Mazza G. Lateralisation effect in comprehension of emotional facial expression: a comparison between EEG alpha band power and behavioural inhibition (BIS) and activation (BAS) systems. Laterality 2010;15:361-84.
Ogata S. Human EEG responses to classical music and simulated white noise: effects of a musical loudness component on consciousness. Percept Mot Skills 1995;80:779-90.
Di G-Q., Wu S-X. Emotion recognition from sound stimuli based on back-propagation neural networks and electroencephalograms. J Acoust Soc Am 2015;138:994-1002.
Di GQ, Fan MC, Lin QH. An experimental study on EEG characteristics induced by intermittent pure tone stimuli at different frequencies. Appl Acoust 2018;141:46-53.
Koelstra S, Mühl C, Soleymani M et al.
DEAP: a database for emotion analysis; using physiological signals. IEEE Trans Affect Comput 2012;3:18-31.
Liu J, Meng H, Nandi A, Li M. “Emotion detection from EEG recordings.” 2016 12th International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery (ICNC-FSKD), 2016. pp. 1722-1727.
Al-Nafjan A, Hosny M, Al-Ohali Y, Al-Wabil A. Review and classification of emotion recognition based on EEG brain-computer interface system research: a systematic review. Appl Sci 2017;7:1239.
Anderson EW, Potter KC, Matzen LE, Shepherd JF, Preston GA, Silva CT. A user study of visualization effectiveness using EEG and cognitive load. Comp Graph Forum 2011;30:791-800.
Badcock NA, Mousikou P, Mahajan Y, De Lissa P, Thie J, McArthur G. Validation of the Emotiv EPOC® EEG gaming systemfor measuring research quality auditory ERPs. PeerJ 2013;2013:1-17.
Schatz R, Sackl A, Timmerer C, Gardlo B. Towards subjective quality of experience assessment for omnidirectional video streaming. 2017 9th International Conference on Quality of Multimedia Experience, QoMEX 2017. pp. 1-6.
Li H, Lau SK. A review of audio-visual interaction on soundscape assessment in urban built environments. Appl Acoust 2020;166:107372.
Hamada M, Zaidan BB, Zaidan AA. A systematic review for human eeg brain signals based emotion classification, feature extraction, brain condition, group comparison. J Med Syst 2018;42:162.
Suhaimi NS, Mountstephens J, Teo J. EEG-Based Emotion Recognition: A State-of-the-Art Review of Current Trends and Opportunities. Comput Intell Neurosci 2020;2020:887542.
Homan RW, Herman J, Purdy P. Cerebral location of international 10–20 system electrode placement. Electroencephalogr Clin Neurophysiol 1987;66:376-82.
Gustafsson F. Determining the initial states in forward-backward filtering. IEEE Trans Signal Process 1996;44:988-92.
Hyvarinen A. Fast and robust fixed-point algorithms for independent component analysis. IEEE Trans Neural Networks 2008;10:626-34.
Calude CS, Dinneen MJ, Pǎun G, Pérez Jiménez MJ, Rozenberg G. Independent Component Analysis and Blind Signal Separation, vol. 3699;2005.
Varinen AH. Fast and robust fixed-point algorithms for independent component analysis. Neural Networks, IEEE Trans 2008; 2008:10.
Salem O, Naseem A, Mehaoua A, “Ep ileptic seizure detection from EEG signal using Discrete Wavelet Transform and Ant Colony classifier,” 2014 IEEE International Conference on Communications (ICC), 2014, pp. 3529-3534 Salem O, Naseem A, Mehaoua A. Epileptic seizure detection from EEG signal using Discrete Wavelet Transform and Ant Colony classifier. 2014 IEEE International Conference on Communications, ICC; 2014. p 3529-34. https://doi.org/10.1109/ICC 2014. 6883868
Sartoretto F, Ermani M. Automatic detection of epileptiform activity by single-level wavelet analysis. Clin Neurophysiol 1999;110:239-49.
Wang Y, Sokhadze EM, El-Baz AS et al.
Relative power of specific eeg bands and their ratios during neurofeedback training in children with autism spectrum disorder. Front Hum Neurosci 2016;9:723.
Ko KE, Yang HC, Sim KB. Emotion recognition using EEG signals with relative power values and Bayesian network. Int J Control Autom Syst 2009;7:865-70.
Pfurtscheller G, Aranibar A. Evaluation of event-related desynchronization (ERD) preceding and following voluntary self-paced movement. Electroencephalogr Clin Neurophysiol 1979;46:138-46.
McDonald JJ, Störmer VS, Martinez A, Feng W, Hillyard SA. Salient sounds activate human visual cortex automatically. J Neurosci 2013;33:9194-201.
Chayer C, Freedman M. Frontal lobe functions. Curr Neurol Neurosci Rep 2001; 1: 547–52.
Stuss DT. Functions of the frontal lobes: relation to executive functions. J Int Neuropsychol Soc 2011;17:759-65.
Ahernt GL, Schwartz GE. Differential lateralization for positive and negative emotion in the human brain: EEG spectral analysis*. Neuropsychologia 1985;23:745-55.
Ma Q, Shang Q, Fu H, Chen F. Mental workload analysis during the production process: EEG and GSR activity. Appl Mech Mater 2012;220-223:193-7.
Wiwatwongwana D, Vichitvejpaisal P, Thaikruea L, Klaphajone J, Tantong A, Wiwatwongwana A. The effect of music with and without binaural beat audio on operative anxiety in patients undergoing cataract surgery: a randomized controlled trial. Eye (Basingstoke) 2016;30:1407-14.
Teplan M, Krakovská A, Štolc S. EEG responses to long-term audio-visual stimulation. Int J Psychophysiol 2006;59:81-90.
Quaedflieg CWEM, Smulders FTY, Meyer T, Peeters F, Merckelbach H, Smeets T. The validity of individual frontal alpha asymmetry EEG neurofeedback. Soc Cogn Affect Neurosci 2016;11:33-43.
Gable PA, Poole BD. Influence of trait behavioral inhibition and behavioral approach motivation systems on the LPP and frontal asymmetry to anger pictures. Soc Cogn Affect Neurosci 2014;9:182-90.
Missonnier P, Deiber MP, Gold G et al.
Frontal theta event-related synchronization: comparison of directed attention and working memory load effects. J Neural Transm 2006;113:1477-86.
Choi Y, Kim M, Chun C. Measurement of occupants’ stress based on electroencephalograms (EEG) in twelve combined environments. Build Environ 2015;88:65-72.
Schadow J, Lenz D, Thaerig S, Busch NA, Fründ I, Herrmann CS. Stimulus intensity affects early sensory processing: sound intensity modulates auditory evoked gamma-band activity in human EEG. Int J Psychophysiol 2007;65:152-61.
Aletta F, Kang J, Astolfi A, Fuda S. Differences in soundscape appreciation of walking sounds from different footpath materials in urban parks. Sustain Cities Soc 2016;27:367-76.
Axelsson Ö, Nilsson ME, Berglund B. A principal components model of soundscape perceptiona. J Acoust Soc Am 2010;128:2836.
Knight RT, Scabini D, Woods DL, Clayworth CC. Contributions of temporal-parietal junction to the human auditory P3. Brain Res 1989;502:109-16.
Jensen O, Tesche CD. Frontal theta activity in humans increases with memory load in a working memory task. Eur J Neurosci 2002;15:1395-9.
Wascher E, Rasch B, Sänger J et al.
Frontal theta activity reflects distinct aspects of mental fatigue. Biol Psychol 2014;96:57-65.
Mikutta C, Altorfer A, Strik W, Koenig T. Emotions, arousal, and frontal alpha rhythm asymmetry during beethoven’s 5th symphony. Brain Topogr 2012;25:423-30.
Vaitl D, Gruzelier J, Jamieson GA et al.
Psychobiology of altered states of consciousness. Psychol Bull 2005;131:98-127.
Doesburg SM, Roggeveen AB, Kitajo K, Ward LM. Large-scale gamma-band phase synchronization and selective attention. Cerebral Cortex 2008;18:386-96.
Jafari MJ, Khosrowabadi R, Khodakarim S, Mohammadian F. The effect of noise exposure on cognitive performance and brain activity patterns. Open Access Maced J Med Sci 2019;7:2924-31.
Balasubramanian G, Kanagasabai A, Jagannath M, Seshadri NPG. Music induced emotion using wavelet packet decomposition—an EEG study. Biomed Signal Process Control 2018;42:115-28.
Department of Architecture and Planning, Indian Institute of Technology, Roorkee, 247667
Source of Support: None, Conflict of Interest: None
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6], [Figure 7], [Figure 8], [Figure 9], [Figure 10]
[Table 1], [Table 2], [Table 3], [Table 4], [Table 5], [Table 6]