Emotion Classification Through Speech Recognition Using LSTM Model and MFCC Feature Extraction

In an era where human-machine interaction is becoming increasingly important, the ability to understand emotions can enhance user experience across various applications, such as customer service and monitoring systems. This study aims to develop an emotion recognition system through voice using the Long Short-Term Memory (LSTM) model and Mel-Frequency Cepstral Coefficients (MFCC) feature extraction. The dataset used in this research includes seven emotion classes: happy, sad, angry, surprised, neutral, fearful, and disgusted, with a total of 2.800 samples obtained from the open-source Kaggle platform. The LSTM model is implemented to capture temporal patterns in voice data, while MFCC features are employed to represent the characteristics of the voice. The evaluation results show a classification accuracy of 98.80%, with the LSTM model proving effective in addressing the challenges of time series data with long-term dependencies. This research not only contributes to the understanding of emotions through voice but also opens opportunities for further development of real-time emotion recognition systems. The implementation of this technology can enrich human-machine interaction, creating more responsive and emotionally aware systems, and contribute to innovations in artificial intelligence.