Research on Lightweight Speaker Recognition Algorithm Based on Attention Mechanism

A lightweight speaker recognition algorithm based on attention mechanism is proposed to address the issue of poor performance and lack of robustness in lightweight speaker recognition models, achieving a lightweight end-to-end speaker recognition model that is robust to noise. On the TDNN model, deep separable convolution is used for lightweight, and convolutional block lightweight convolutional block attention module (CBAM) attention module is used to pay attention to the channel and spatial dimensions of features, making the model more focused on the speaker's personality characteristics. Data augmentation uses speech data denoising and SpecAugmention Mel spectrum enhancement to achieve lightweight while also possessing robustness. The experimental results show that the proposed CBAM-TDNNLite model has the least number of parameters, but its performance is the best. The model not only saves computational costs, but also has good robustness against noise interference.