EEG based brain signals analysis (scent classification)
Over the years, Electroencephalogram (EEG) brain signals have been found closely related to human’s physical and biological activities. These include mechanical moves, emotional states, thoughts and the perceiving of external stimuli. A final year project has been conducted to collect and compare hu...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project (FYP) |
| Language: | English |
| Published: |
2013
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/10356/54478 |
| _version_ | 1826115186278268928 |
|---|---|
| author | Yang, Jiong |
| author2 | Ser Wee |
| author_facet | Ser Wee Yang, Jiong |
| author_sort | Yang, Jiong |
| collection | NTU |
| description | Over the years, Electroencephalogram (EEG) brain signals have been found closely related to human’s physical and biological activities. These include mechanical moves, emotional states, thoughts and the perceiving of external stimuli. A final year project has been conducted to collect and compare human brain signals between relaxation and scent stimulation and between different scent stimulations as well. Lavender and peppermint scents were used in this project. This report aims to present in detail the design and conduction of the EEG wave collection experiment and the analyses on the collected signals.
20 participants joined the experiment for data collection and the brain waves were collected by an EEG system in a tightly controlled environment. Multiple electrodes were placed at different regions (different lobes) of the brain. Essential oils and diffuser were used to generate the scent stimulation.
Both time domain and frequency domain features were extracted from the signals. These features included the time domain statistics, frequency domain statistics, band powers and discrete wavelet coefficients. The first three sets of features were evaluated individually by direct feature value comparison before and after scent stimulation and K-Mean Clustering. The highest clustering accuracy for K-Mean was 61.5% and the high frequency band (Upper Beta and Gamma) power feature outperformed the rest. The fourth feature was evaluated by support vector machines with RBF kernel. The highest classification accuracy was 61.9% and the wavelet coefficients at the resolution of 16-32Hz outperformed the rest. |
| first_indexed | 2024-10-01T03:51:05Z |
| format | Final Year Project (FYP) |
| id | ntu-10356/54478 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T03:51:05Z |
| publishDate | 2013 |
| record_format | dspace |
| spelling | ntu-10356/544782023-07-07T16:19:33Z EEG based brain signals analysis (scent classification) Yang, Jiong Ser Wee School of Electrical and Electronic Engineering DRNTU::Engineering Over the years, Electroencephalogram (EEG) brain signals have been found closely related to human’s physical and biological activities. These include mechanical moves, emotional states, thoughts and the perceiving of external stimuli. A final year project has been conducted to collect and compare human brain signals between relaxation and scent stimulation and between different scent stimulations as well. Lavender and peppermint scents were used in this project. This report aims to present in detail the design and conduction of the EEG wave collection experiment and the analyses on the collected signals. 20 participants joined the experiment for data collection and the brain waves were collected by an EEG system in a tightly controlled environment. Multiple electrodes were placed at different regions (different lobes) of the brain. Essential oils and diffuser were used to generate the scent stimulation. Both time domain and frequency domain features were extracted from the signals. These features included the time domain statistics, frequency domain statistics, band powers and discrete wavelet coefficients. The first three sets of features were evaluated individually by direct feature value comparison before and after scent stimulation and K-Mean Clustering. The highest clustering accuracy for K-Mean was 61.5% and the high frequency band (Upper Beta and Gamma) power feature outperformed the rest. The fourth feature was evaluated by support vector machines with RBF kernel. The highest classification accuracy was 61.9% and the wavelet coefficients at the resolution of 16-32Hz outperformed the rest. Bachelor of Engineering 2013-06-21T01:38:46Z 2013-06-21T01:38:46Z 2013 2013 Final Year Project (FYP) http://hdl.handle.net/10356/54478 en Nanyang Technological University 100 p. application/pdf |
| spellingShingle | DRNTU::Engineering Yang, Jiong EEG based brain signals analysis (scent classification) |
| title | EEG based brain signals analysis (scent classification) |
| title_full | EEG based brain signals analysis (scent classification) |
| title_fullStr | EEG based brain signals analysis (scent classification) |
| title_full_unstemmed | EEG based brain signals analysis (scent classification) |
| title_short | EEG based brain signals analysis (scent classification) |
| title_sort | eeg based brain signals analysis scent classification |
| topic | DRNTU::Engineering |
| url | http://hdl.handle.net/10356/54478 |
| work_keys_str_mv | AT yangjiong eegbasedbrainsignalsanalysisscentclassification |