A Multi-Modal Deep Learning Approach for Emotion Recognition

In recent years, research on facial expression recognition (FER) under mask is trending. Wearing a for protection from Covid 19 has become compulsion and it hides the expressions that why FER difficult task. The prevailing unimodal techniques are not up to mark in terms of good results masked face, however, multimodal technique can be employed generate better results. We proposed methodology based deep learning face using vocal expressions. been trained dataset. have used two standard datasets, M-LFW dataset CREMA-D TESS form audio while faces data image heterogenous. order make homogeneous, voice converted into images by taking spectrogram. A spectrogram embeds important features converts format images. Later, passed training. neural network experimental demonstrate algorithm outsets methods other state-of-the-art models.