Learning content‐adaptive feature pooling for facial depression recognition in videos

Abstract

Recently, a deep representation of facial depression built on convolutional neural networks has shown impressive performance in videobased depression recognition. However, most existing approaches either fix the weights or using a certain heuristics to integrate the frame-level facial features, resulting in suboptimal feature aggregation in encoding the helpful while discarding noisy information in videos. To address this issue, the authors introduce the memory attention mechanism in a regression network to learn a deep discriminative depression representation, where the residual network module aims at learning frame-level deep feature, while the attention module acts as a pooling layer by adaptively learning the weights emphasising or suppressing face images with varying poses and imaging conditions. They empirically evaluate the proposed approach on a benchmark depression dataset, and the results demonstrate the superiority of their approach over the state-of-the-art alternatives.

Encoding Temporal Information For Automatic Depression Recognition From Facial Analysis

“…Our proposed method achieves better results than the schemes in [14,23,24] which are based on handcrafted features. The proposed method also outperforms the deep learning schemes proposed in [21,3,11,22,9] on AVEC2014, confirming the good performance of our model. In [25], the method is based on distribution learning with expectation loss function.…”

“…In addition, our proposed method achieves better results than the method in [9] which uses a four-stream model to explore multiple facial regions, showing the importance of exploring the temporal information for depression detection. Finally, the authors in [22] explore spatial information and employ attention mechanism to fuse facial features. Our proposed method outperforms such method, which suggest that exploring temporal information between the frames is a better approach.…”

“…Zhou et al [9] used multiple 2D CNNs to investigate different facial regions with a scheme to combine all responses. In [10], the authors proposed to integrate the facial features by using an attention mechanism to fuse features from a video. Such approaches consider the spatial dependencies of extracted features, disregarding the temporal information between the frames, which can impair the exploitation of spatio-temporal relations.…”

A review and meta-analysis of machine intelligence approaches for mental health issues and depression detection

“…Mean accuracy achieved was 86.6% and maximum accuracy achieved was 86.9% Zhou et al (2019) [146] T his paper studies the facial expression of the videos. T hey took the content from the adaptive feature by extracting.…”

Identifying Psychological Symptoms Based on Facial Movements

“…Mental illnesses have a significant impact on an individual's physical health (1), achievements (2,3), and life satisfaction (4). In addition to scales, behavioral recognition methods have been developed to judge the existence (5) or degree (6,7) of specific mental illnesses. However, identifying an individual's mental health status from a range of perspectives may be more helpful in non-professional scenarios such as self-monitoring or large-scale monitoring.…”