Modified Squeezenet for Speech Emotion Recognition with Extractive Feature Set

SER is crucial for human-machine interaction applications, necessitating efficient methods to discern emotional states from spoken language. This paper introduces a novel SER approach consisting of three main steps: preprocessing, feature extraction, and emotion recognition. In the preprocessing phase, the input speech signal undergoes initial processing to enhance its suitability for subsequent analysis. Following this, relevant acoustic features such as MFCC, spectral features, and zero-crossing rate are extracted from the preprocessed signal. Subsequently, a modified Squeezenet architecture is proposed for emotion recognition, leveraging the extracted feature set. Squeezenet, known for its compact design and efficient resource utilization, is adapted to discern emotional states from speech signals. Through experimental validation on benchmark datasets, the efficacy of the proposed approach in achieving state-of-the-art performance in speech emotion recognition tasks is demonstrated. This method enhances the accuracy and efficiency of emotion recognition and holds promise for practical applications in real-world scenarios, such as emotion-aware human-computer interfaces and virtual assistants.