Depression is a common, but serious mental disorder that affects people all over the world. Therefore, an automatic depression assessment system is demanded to make the diagnosis of this disorder more popular. We propose a multimodal fusion of speech and linguistic representations for depression detection. We train our model to infer the Patient Health Questionnaire (PHQ) score of subjects from the E-DAIC corpus. For the speech modality, we apply VGG-16 extracted features to a Gated Convolutional Neural Network (GCNN) and a LSTM layer. For the linguistic representation, we extract BERT features from transcriptions and input them to a CNN and a LSTM layer. We evaluated the feature fusion model with the Concordance Correlation Coefficient (CCC), achieving a score of 0.696 on the development set and 0.403 on the testing set. The inclusion of visual features is also discussed. The results of this work were submitted to the Audio/Visual Emotion Challenge and Workshop (AVEC 2019).
各種検索
サポート
ヘルプ