Metaheuristically Enabled System for Emotion Recognition using BiLSTM

Emotion recognition from speech is vital for applications in human-computer interaction and mental health diagnostics. This paper presents an efficient approach to classify emotions using the Toronto Emotional Speech Set (TESS) dataset. Key audio features, including Mel-Frequency Cepstral Coefficients (MFCCs) and spectral characteristics, are extracted to represent the speech signals. To enhance computational efficiency and mitigate overfitting, metaheuristic optimization techniques, such as Genetic Algorithms (GA) and Particle Swarm Optimization (PSO), are utilized for dimensionality reduction by selecting the most relevant features. These optimized features are then fed into a Bidirectional Long Short-Term Memory (BiLSTM) network, which effectively captures temporal dependencies in the speech data. The proposed system combines the strength of metaheuristic feature selection with the powerful learning capabilities of BiLSTM, achieving superior classification accuracy compared to traditional methods. Experimental results validate the efficacy of this hybrid approach, offering a robust and scalable solution for emotion recognition tasks. Keywords: Speech Emotion Recognition, TESS Dataset, Metaheuristic Optimization, Dimensionality Reduction, BiLSTM, Deep Learning