Title：Low-power-consumption analog domain feature vector extraction method for voiceprint recognition
Country：China
Patent No.：202010577295.4
Legal Status：Authorized
Inventor：Ning Pu, Hanjun Jiang, Chun Zhang, Zhihua Wang
Assignee：Tsinghua University
Address：Tsinghua University,Haidian District Beijing 100084, China
Filing Date：2020-06-22
Issue Date：2022-10-14
Abstract：
The invention relates to a low-power-consumption analog domain feature vector extraction method for voiceprint recognition. Voiceprint recognition is completed by adopting a mixed domain architecturecombining analog domain feature extraction and a digital domain recognition model. Because classic digital domain MFCC feature extraction contains operations such as FFT, DCT and the like, a large amount of power consumption is consumed, and the requirements of a voice equipment awakening circuit in a continuous working state on low power consumption and low calculated amount are not met. The low-dimension and low-hardware-overhead analog domain speech feature extraction method provided by the invention can be specifically extended into two modes of full-analog filtering feature extraction andmixed feature extraction, and is respectively suitable for application scenes with relatively high requirements on recognition accuracy and strict requirements on ultra-low power consumption limitation.
Patent Certificate：
PDF/Jpg