Real Time Speech Emotion Recognition for Mental Health Monitoring System

With the rise in mental health disorders, there’s a need for new monitoring tools. This paper introduces a real-time speech emotion recognition (SER) model that uses AI to detect emotions like happiness, sadness, anger, fear, and neutrality from speech. The model combines deep learning with features such as Mel-frequency cepstral coefficient (MFCCs), pitch, and energy. It employs convolutional and recurrent neural networks and improves accuracy through transfer learning with a pre-trained emotional speech dataset. Optimized for speed and efficiency, the model operates on edge devices for mobile and wearable health apps, allowing for continuous monitoring and immediate feedback. It also addresses potential biases and is tested in real-world scenarios for robustness and adaptability. Ongoing enhancements and ethical considerations aim to maximize its effectiveness in mental health care.