Physiological Signal Based Depression Detection System
According to World Health Organization (WHO) almost 300 million people are suffering from depression that ultimately results in forlorn, and desperation. According to researchers, main reason of depression is that people meditate beyond their scope and ultimately draw the false conclusion and they b
2025-06-28 16:34:28 - Adil Khan
Physiological Signal Based Depression Detection System
Project Area of Specialization Biomedical EngineeringProject SummaryAccording to World Health Organization (WHO) almost 300 million people are suffering from depression that ultimately results in forlorn, and desperation. According to researchers, main reason of depression is that people meditate beyond their scope and ultimately draw the false conclusion and they become sad. But more harrowing is that almost 80,0000 people die every year in the world. However, treatments are available to get someone extricated from depression, but these are human intensive and always need a doctor. People show reluctance to visit psychiatrists due to fear of social stigma. Depression detection using physiological signals like Electroencephalogram (EEG), Galvanic Skin Response (GSR), and Photoplethysmogram (PPG) is quite useful technique to diagnose the depression. The diagnosis of depression in early curable stages is critical and might save the life of a patient. Machine learning algorithms will use to extract meaningful information from raw data. Any depression problem can be tackled, mental reactive and proactive disturbance can be dealt properly through this system. Cure of depression through pervasive methods has been top preference of researchers for last ten years. Different researchers have shown various results pertaining to accuracy. They did research by reducing, increasing number of patients and obtained accuracy at different levels. Three electrode-based depression detection method is easy and fast as well ensures data accuracy. Absolute power of theta wave is the feature which determines the performance of best algorithm. It displays connection between power of theta wave and depression.
In our project, an experimental protocol will be designed in which negative, positive, and neutral stimulus in the form of images are shown to subjects and their physiological signals like EEG, PPG, GSR, and heart sounds will be recorded. EEG signal are used to detect brain activity, but it is mixed with noise and interferences due to power line and can have a bad impact on the accuracy of the system. Electromyography (EMG) and Electrooculogram (EOG) signals can also be recorded with the help of EEG sensors. So, in order to get accurate result in the phase of feature extraction and classification the data in raw form should be denoised first.
Construction of feature matrix n x m. With the change in emotional state the feature of EEG also changes accordingly. Following features were selected for extraction:
Selecting relevant features and discarding others is a very important step in classification problem. We will use SVM, KNN and CT classification algorithm for data processing as our study shows that this algorithm has highest accuracy with EEG based depression detection system .We use EEG, Heart Sound, PPG and GSR sensors to measure all the aspects of mentally disturbed and depressed person.
Project ObjectivesThe objectives set for this project are
- To acquire physiological data in rest, positive, negative and neutral stage by showing the subject positive, negative and neutral images respectively.
- To extract features from acquired data. Such as time domain features that are extracted in this case are peak, variance, skewness, kurtosis and Hjorth parameter and frequency domain features that are relative centroid frequency, absolute centroid frequency relative power, and absolute power
- To detect different depression states using classifiers. We will use SVM, KNN and CT classification algorithm for data processing as our study shows that this algorithm has highest accuracy with EEG based depression detection system.
Proposed Methodology
Recording as well as analyzing EEG, ECG, PPG, GSR signals and heart sounds of the patient in different stimuli may help identify that whether person is depressed or not. In the above discussed model, we are recording EEG signal of the patient is different states that are described below:
- Rest State.
- Positive Stimuli.
- Negative Stimuli.
- Neutral Stimuli
Before starting our experiment, we need some prerequisites that must be followed in order to start our experiment in a legal way, they are as follow
This form is signed by the participant in order accept any kind of risk that might be involved in the experimental procedure
This form includes all kind of personal information of an individual that is the subject in the experiment
Patient whose signal is to be recorded was made to view positive, negative and neutral images.
Experiment took place in complete silence . All the credentials requirement was explained to the participant. The Emotiv device with five channels was placed on the head of the participant and all settings were configured. In first phase data was recorded in rest position of the participant whose duration was 120s and after first phase there was a 60s break and after which second phase begins in which participant was made to see images for 120s followed by a 60s break which continues until the experiment is completed.
Above experiment comprises of three types of images, which are categorized as follow:
- Positive.
- Negative.
- Neutral.
For the purpose of cross-referencing following forms were filled by the participant:
EEG signal are used to detect brainwave activity, but it is mixed with noise and interferences due to power line and can have a bad impact on the accuracy of the system. ECG, EMG and EOG signals can also be recorded with the help of EEG sensors. So, in order to get accurate result in the phase of feature extraction and classification the data in raw form should be denoised first.
Feature Matrix Construction
Construction of feature matrix n x m involves following steps:
- Identify and extract all the features for each set of EEG data.
- Identify the most suitable features from all the extracted features to form a final feature vector.
- Tag each row as depressed or not depressed.
With the change in emotional state the feature of EEG also changes accordingly. Following features were selected for extraction:
- Time Domain Features.
- Frequency Domain Features.
Feature Selection
Selecting relevant features and discarding others is a very important step in classification problem. The minimal-redundancy-maximal-relevance is a technique to perform feature-selection.
Classification
We will use SVM, KNN and CT classification algorithm for data processing as our study shows that this algorithm has highest accuracy with EEG based depression detection system.
Benefits of the ProjectThe project is innovative alternative to traditional depression detection techniques. It will assist to detect depression among affected patients quite accurately and easily. Application of computing technology in medical has revolutionized it. Physiological Signal Based Depression Detection is one of those pervasive techniques that can replace the erroneous and traditional depression detection methods perennially.
Technical Details of Final DeliverableOur deliverable was to determine the state in which depression can be detected more accurately. The Emotiv device with five channels recorded EEG signal of the subject. Shimmer kit that recorded GSR and PPG signal of subject, worn on the hand of subject. Subject whose signal is to be recorded was made to see images. Our algorithm determined whether the subject is depressed or not on the basis of acquired physiological signals.
Final step was to upload acquired data to the trained model as described in project implementation method which will classify the subject in one of the following categories:
- Normal.
- Depressed.
| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 74452 | |||
| Rechargeable Power Cell | Equipment | 4 | 340 | 1360 |
| Rechargeable Power Cell Device | Equipment | 2 | 950 | 1900 |
| Shimmer optical pulse-ear clip | Equipment | 2 | 4241 | 8482 |
| Muse | Equipment | 1 | 33500 | 33500 |
| ECG electrode | Equipment | 150 | 40 | 6000 |
| Print outs | Miscellaneous | 1250 | 4 | 5000 |
| 9 In biophysical lid pack | Equipment | 880 | 12 | 10560 |
| Optical Pulse Probe | Equipment | 1 | 7650 | 7650 |