A Parallel-Model Speech Emotion Recognition Network Based on Feature Clustering

Speech Emotion Recognition (SER) is a common aspect of human-computer interaction and has significant applications in fields such as healthcare, education, elder care. Although researchers have made progress speech emotion feature extraction model identification, they struggled to create an SER system with satisfactory recognition accuracy. To address this issue, we proposed novel algorithm called F-Emotion select features established parallel deep learning recognize different types emotions. We first extracted from calculated the value for each feature. These values were then used determine combination that was optimal recognition. Next, input train test type emotion. Finally, decision fusion applied output results obtain overall result. analyses conducted on two datasets, RAVDESS EMO-DB, accuracy reaching 82.3% 88.8%, respectively. The demonstrate can effectively analyze correspondence between types.The MFCC best describes emotions Neutral, Happiness, Fear Surprise, Mel Angry Sadness.The mechanism improve