Voice print detection method, device and equipment and storage medium based on short texts

Abstract

The invention discloses a voice print detection method, device and equipment and a storage medium based on short texts. The voice print detection method includes the steps of training a preset deep neural network by adopting training samples; acquiring speech signals to be recognized; preprocessing the speech signals to be recognized, and carrying out feature extraction on the preprocessed speechsignals to obtain Mel frequency cepstrum coefficients; transmitting the Mel frequency cepstrum coefficients to the pre-trained deep neural network as input vectors,, obtaining output vectors of the deep neural network on a last full connection layer and using the output vectors as voice print vectors of the speech signals; and comparing the voice print vectors of the speech signals with pre-storedvoice print vectors in a voice print model base, and outputting voice print detection results according to compared results, wherein the training samples and the speech signals both are the short texts. The voice print detection method solves the problems of long speech signals, large sample information quantities and high requirements for computing resources in existing voice print detection methods.

Description

technical field [0001] The present invention relates to the field of information technology, in particular to a voiceprint detection method, device, equipment and storage medium based on short text. Background technique [0002] Voiceprint detection is a common and effective identification method, which can be applied to a series of scenarios that need to be combined with identity authentication, such as network payment, voiceprint lock control, survival authentication, and Internet of Things device verification, especially when it is inconvenient to use video image verification In the remote verification of , it is not limited by the device at all. When performing verification, using content and voiceprint detection for double verification can greatly increase the threshold of being attacked and improve security. When performing voiceprint detection, currently commonly used methods include but are not limited to template matching methods, probability model methods, artific...

AI Technical Summary

Problems solved by technology

[0003] The embodiment of the present invention provides a short text-based voiceprint detection method, device, device and storage medium to solve the problems of redundant voice signals, large amount of sample information, and high requirements for computing resources in existing voiceprint detection methods

Images