Abnormal sound quality data detection method, device, electronic device and storage medium

Abstract

Embodiments of the present application provide a method, device, electronic device, and storage medium for detecting abnormal sound quality data. The method includes: receiving user voice data, determining voiced-segment voice data and non-voice-segment voice data in the user voice data; extracting voiced-segment voice data The acoustic features of the voiced segment in the data are extracted, and the acoustic features of the voiceless segment in the speech data of the silent segment are extracted; the acoustic features of the voiced segment are input into the trained voiced segment production model to obtain the voiced segment detection results of the user's voice data, and the acoustical features of the silent segment are extracted. Input the trained silent segment generation model to obtain the silent segment detection result of the user's speech data; the voiced segment generation model is trained with normal voiced segment samples, and the silent segment generation model is trained with normal silent segment samples; according to the voiced segment detection The result and the silent segment detection result determine whether the user speech data is abnormal sound quality data. The embodiment of the present application can improve the accuracy rate of abnormal sound quality data detection.

Description

technical field [0001] The present application relates to the technical field of machine learning, and in particular, to a method, device, electronic device and storage medium for detecting abnormal sound quality data. Background technique [0002] In the voice evaluation system, if the evaluation equipment fails, the evaluation results of the voice evaluation system are unsatisfactory. Therefore, there is a need for a system for identifying abnormal sound quality data, which can identify whether voice data has abnormal sound quality. [0003] The current process of detecting abnormal sound quality data is as follows: in the training phase, the positive example data (normal sound quality data) and the negative example data (the sound quality abnormal data) are prepared in advance, and the classification model is trained by using the positive example data and the negative example data; The corresponding features are extracted from the speech data to be tested, and the classi...

AI Technical Summary

Problems solved by technology

However, in practical applications, abnormal data is ever-changing, and it is difficult to list them completely, which leads to error-prone judgments on abnormal data.

Images