Automatic Emotion Recognition System using tinyML

This paper proposes a system to detect and analyze emotions using tinyml. Automated emotion detec-tion has attracted increasing interest from researchers in neuroscience, psychology, computer science, and associated disciplines for the past 3 decades. In this paper, an Automatic Emotion Recognition System implemented using tinyml (AERSUT) has been proposed. There are 3 levels/stages for the above-proposed method -1) input data using tinyml, 2) feature extraction, 3) emotion recognition. The first section involves giving input raw speech data to analyze and detect through a tinyml board. Features like Mel Frequency Cepstral Coefficients(MFCC), root mean square energy, and zero crossing rate are extracted from this audio and the user’s emotional state is detected. Here we classify emotions into 8 categories-surprise, neutral, disgust fear, sad, calm, happy, and anger. 2 models were trained using TensorFlow and the model with the best accuracy has been used for detection.