SVM Speech Emotion Recognition Based on Matlab GUI
Speech Emotion Recognition (SER) is to judge the emotional state expressed by the speaker by analyzing and processing the speech signal, extracting relevant features and emotional information from it. Voice emotion recognition has been widely used in the fields of emotional intelligence, intelligent customer service, and auxiliary medical treatment. SVM (Support Vector Machine) is a common machine learning method, often used in classification problems. This article will introduce the implementation process of SVM speech emotion recognition based on Matlab GUI, and provide the corresponding source code.
1. Dataset
When performing speech emotion recognition, it is necessary to use speech datasets for model training and testing. The data set used in this paper is the RAVDESS (Ryerson Audio-Visual Database of Emotional Speech and Song) emotional voice database. This data set contains voice audio files from 24 actors in different emotional states. It contains a total of 8 emotions, respectively For Angry, Disgusted, Feared, Happy, Sad, Surprised, Neutral and Angry. Each speech audio file is sampled at 48kHz. For the convenience of data preprocessing, we can reduce the sampling frequency of the original audio file to 16kHz.
2. Feature extraction
During model training, it is necessary to convert speech signals into digital feature vectors for processing. Common feature extraction methods include MFCC (Mel-Frequency Cepstral Coefficients), LPC (Linear Predictive Coding), LPCC (Linear Predictive Cepstral Coefficients), etc. In this paper we use MFCC as the feature extraction method.
The MFCC feature extraction function mfcc.m is provided in Matlab, which can easily calculate the MFCC coefficients. Here is the corresponding code: