Voiceprint recognition method under channel attention propagation and aggregation

Abstract

The invention relates to a voiceprint recognition method under channel attention propagation and aggregation, belonging to the field of signal processing. The method comprises the following steps: S1, carrying out second-order wavelet scattering transformation on an original speech discrete signal; S2, performing voiceprint mapping coding of multi-scale features; and S3, evaluating the similarity of the voiceprint codes. According to the method, the multi-scale short-time voice features are obtained through wavelet scattering transformation, and the multi-scale features are mapped by adopting the time delay neural network based on channel attention propagation and aggregation to obtain the voiceprint codes, so the accuracy and robustness of voiceprint recognition are improved. The method takes the processing of long-time and short-time voices into consideration, provides a new technical means for voiceprint recognition containing short-time voice data, and can also be migrated to other voice processing fields to serve as one of voiceprint code acquisition methods.

Description

technical field [0001] The invention belongs to the field of signal processing, and relates to a voiceprint recognition method under channel attention propagation and aggregation. Background technique [0002] As a biometric technology, voiceprint recognition has the following advantages compared with face recognition, fingerprint recognition and other technologies: (1) easy to obtain; (2) low cost; (3) high user acceptance; (4) universal wide range. In recent years, the research on using the hidden layer output of neural network to encode voiceprint has made remarkable progress. However, due to the few voiceprint features extracted from the voiceprint data containing short-term speech and the poor robustness of the voiceprint features, the reliable operation of the voiceprint recognition system still faces major challenges. [0003] Many research works use data sets such as Voxceleb or Librispeech for modeling and verification. The average audio duration of these data set...

AI Technical Summary

Problems solved by technology

The existing models only extract various attribute features of the speaker’s voiceprint from the frame-level features of the last layer of the network, and do not pay attention to the frame-level features extracted by other network layers and the rich information contained in each channel, that is, The useful information of voiceprint features is not captured and emphasized, which makes the voiceprint encoding output by the network less robust and reduces the performance of voiceprint recognition

Images