An Emotion Recognition from Speech using LSTM

Résumé fourni par la source

The spoken emotions of people are frequently not recognized by machine learning algorithms. Applications that analyze voice emotions in real-time heavily rely on Speech Emotion Recognition (SER). It can be applied in a variety of situations, including human behavior analyses and emergency centers. A new study topic that has now emerged is the detection and classification of emotions. Previous research has looked at a variety of emotional classification methods. Due to their excellent qualities, speech signals make a wonderful source for computational linguistics. And for this reason, a lot of professionals wish to be able to identify speech emotion. In the past ten years, social interaction has placed a lot of emphasis on how speech expresses emotion. However, due to the lack of information on the basic temporal link of the waveform, the actual effectiveness of identification needs to be improved. A novel approach to voice recognition is currently recommended, integrating structured audio information with long-term neural networks, in order to fully exploit the shift in emotional content over phases (LSTM) To determine emotion concentration in several blocks, a few LSTM-based optimal techniques are provided. By modifying the traditional forgetting gate, the technique initially reduces computation costs. Second, to get task-related information, an attention mechanism is applied to both the time and feature dimensions in the LSTM’s final output rather than using the output from the prior iteration of the conventional method. Furthermore, an efficient method has been employed to locate the spatial and distinctive characteristics in the LSTM’s final output. in order to gather information rather than using the outcomes from the preceding stage of the conventional methodology research.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.