UAED-Abnormal-Emotion-Detection

This repository contains the code for paper "UAED: Unsupervised Abnormal Emotion Detection Network Based on Wearable Mobile Device". The overall structure of the UAED is shown in the figure below.

Implementation

1.Environment

pytorch == 1.5.1
torchvision == 0.6.1
numpy == 1.21.5
scipy == 1.4.1
sklearn == 0.0

2.Dataset

We evaluate the proposed model on four publicly available datasets: (1)DREAMER (2)The Stress Recognition in Automobile Drivers database (DRIVEDB) (3)Mahnob-HCI-tagging database (MAHNOB-HCI) (4)Wearable Stress and Affect Detection (WESAD). Detailed information about the datasets is summarized in the Table below.

3.Pre-processing

We only do simple data cleaning including removal of incomplete data and normalization to better understand the learning ability of UAED for feature representation. We use a stacking operation to transform the original 1D time series into 3D samples, and the visualization from DREAMER is presented as follow.

The top two rows are normal samples and the bottom two rows are abnormal samples.

4.Train and Test the Model

We write both training and evaluation process in the main.py, execute the following command to see the training and evaluation results.
python main.py

Performance

We conduct the experiments under a nested cross-validation leave-one-subject-out (LOSO) procedure. Concretely, we leave the data from one subject at a time as the test set and the rest of the data as the training set, thus avoiding overfitting. Further, we perform a 10-fold cross-validation on the training set as an inner loop. UAED achieve better average results than a lot of existing traditional and deep methods. Here, we only show the F1 score performance, while the other results can be found in the original paper.

Methods	DREAMER	DRIVEDB	MAHNOB-HCI	WESAD	Average
ISF	0.315	0.198	0.682	0.602	0.449
OCSVM	0.128	0.012	0.556	0.658	0.339
AE-LSTM	0.598	0.517	0.620	0.416	0.538
DAGMM	0.412	0.210	0.665	0.523	0.453
DSVDD	0.482	0.472	0.632	0.518	0.526
VAE-LSTM	0.436	0.564	0.606	0.594	0.550
DOMI	0.564	0.552	0.692	0.659	0.617
UAED(Pro.)	0.546	0.612	0.726	0.789	0.668