[Paper Review] Human Emotion Recognition Based On Galvanic Skin Response signal Feature Selection and SVM

Di Fan, Mingyang Liu+2 more|arXiv (Cornell University)|Jul 4, 2023

Emotion and Mood Recognition14 citations

TL;DR

The paper presents a method for human emotion recognition using GSR signals with wavelet denoising, normalization, covariance-based feature selection of 30 features, and SVM classification, achieving accuracy over 66.67%.

ABSTRACT

A novel human emotion recognition method based on automatically selected Galvanic Skin Response (GSR) signal features and SVM is proposed in this paper. GSR signals were acquired by e-Health Sensor Platform V2.0. Then, the data is de-noised by wavelet function and normalized to get rid of the individual difference. 30 features are extracted from the normalized data, however, directly using of these features will lead to a low recognition rate. In order to gain the optimized features, a covariance based feature selection is employed in our method. Finally, a SVM with input of the optimized features is utilized to achieve the human emotion recognition. The experimental results indicate that the proposed method leads to good human emotion recognition, and the recognition accuracy is more than 66.67%.

Motivation & Objective

• Motivate and address emotion recognition from Galvanic Skin Response (GSR) signals.
• Propose a pipeline that denoises, normalizes, and selects features from GSR data.
• Develop an SVM classifier using optimized features for emotion recognition.
• Evaluate the recognition performance on GSR data collected via a wearable platform.

Proposed method

• Acquire GSR data using the e-Health Sensor Platform V2.0.
• Apply wavelet-based denoising to reduce noise in GSR signals.
• Normalize data to mitigate individual differences in GSR.
• Extract 30 features from the normalized signals.
• Use covariance-based feature selection to obtain optimized features.
• Train an SVM classifier on the selected features to recognize emotions.

Experimental results

Research questions

• RQ1 Can GSR signals plus a covariance-based feature selection approach improve emotion recognition accuracy?
• RQ2 What is the impact of feature selection on SVM-based emotion classification performance using GSR data?
• RQ3 How do denoising and normalization affect the recognition results for GSR-based emotion recognition?

Key findings

• The proposed pipeline yields emotion recognition with accuracy greater than 66.67%.
• Covariance-based feature selection reduces feature dimensionality from 30 to optimized features.
• Wavelet denoising and normalization help reduce noise and inter-subject variability in GSR data.
• SVM with optimized features effectively classifies human emotions from GSR signals.