Dialect recognition based on improved GFCC with VMD and Teager energy operator cepstral coefficients

Aiming at the problems of low recognition rate of Chinese dialects and poor robustness in noisy environment, this paper proposes an improved feature extraction algorithm, which combines the Sparrow Search Algorithm (SSA) and Variational Mode Decomposition (VMD) with the improved Gamma-tone Frequency Cepstral Coefficients (GFCC) and Teager Energy Operator Cepstral Coefficient (TEOCC). Firstly, the VMD parameters are optimized by SSA, the dialect signal is decomposed into different modal components, and the correlation coefficients with the original signal are calculated. Then, the component of the dominant signal is screened according to the correlation, which is used to calculate the GFCC characteristics. Subsequently, the improved GFCC features are fused with TEOCC, and Principal Component Analysis (PCA) is used to reduce the dimension of the fusion features. Finally, the Gaussian mixture model-universal background model (GMM-UBM) is used for dialect recognition. The experimental results show that the proposed method can improve the recognition rate compared with the traditional GFCC and single feature in both noise-free and noise-free environments, and verify the effectiveness of the improved algorithm in dialect recognition.