Speech Emotion Recognition using Branched-based Convolutional Neural Network Technique

Speech Emotion Recognition is a significant pattern recognition of human speech using feature extraction for communication media. This paper aims to recognize speech emotion through the CNN branch-based model. Machine learning strategy for reliable speech emotion identification in the face of limited data and cultural differences. This study proposed to analyze three different datasets and observe the most useable emotion accuracy using the CNN model for the best outcome accuracy performance. We highlight the speech operations of RAVDESS, TESS, and SAVEE Datasets using deep learning libraries like Keras-Dense, Conv1D, MaxPooling1D, Flatten, Dropout, and Batch Normalization to build branches of convolutional neural network architectures. The result performance of the model Accuracy is 87.12%. Overall, emotional recognition shows potential for improving human-machine communication and providing insights into specific emotional states.