A Novel Policy for Pre-trained Deep Reinforcement Learning for Speech Emotion Recognition

A Novel Policy for Pre-trained Deep Reinforcement Learning for Speech Emotion Recognition

Deep Reinforcement Learning (deep RL) has gained tremendous success in gaming but it has rarely been explored for Speech Emotion Recognition (SER). In the RL literature, policy used by the RL agent plays a major role in action selection, however, there is no RL policy tailored for SER. Also, an extended learning period is a general challenge for deep RL, which can impact the speed of learning for SER. In this paper, we introduce a novel policy, the 'Zeta policy' tailored for SER and introduce pre-training in deep RL to achieve a faster learning rate. Pre-training with a cross dataset was also studied to discover the feasibility of pre-training the RL agent with a similar dataset in a scenario where real environmental data is not available. We use 'IEMOCAP' and 'SAVEE' datasets for the evaluation with the problem of recognising four emotions, namely happy, sad, angry, and neutral. The experimental results show that the proposed policy performs better than existing policies. Results also support that pre-training can reduce training time and is robust to a cross-corpus scenario.

Files associated with this item cannot be displayed due to copyright restrictions.