Voice section segmentation detection method and system and spoken language detecting and evaluating method and system

Abstract

The invention discloses a voice section segmentation detection method and system. The method includes the steps of obtaining segmentation results obtained by conducting voice section segmentation on reading voices; conducting voice recognition on the reading voices to obtain recognition results; respectively calculating the recognition rate of the segmentation results and the recognition rate of the recognition results according to reading texts corresponding to the reading voices; if the recognition rate of the segmentation results is higher than the recognition rate of the recognition results, determining that the segmentation results are normal; if the recognition rate of the segmentation results is not higher than the recognition rate of the recognition results, determining that the segmentation results are abnormal. The invention further discloses a spoken language detecting and evaluating method and system. By means of the voice section segmentation detection method and system and the spoken language detecting and evaluating method and system, whether the voice section segmentation results are abnormal or not can be effectively detected, and the accuracy of evaluating and detecting results is further guaranteed.

Description

technical field [0001] The invention relates to the technical field of speech signal processing, in particular to a segment segmentation detection method and system, and a spoken language evaluation method and system. Background technique [0002] Spoken language evaluation can realize the evaluation of the speaker's speaking level, and guide the speaker to learn according to the evaluation results. The basic principle of evaluation comes from measuring the difference between standard pronunciation and actual pronunciation. Segment segmentation is a mainstream method in oral language evaluation technology. The specific process is: build a decoding network based on the reading text, and analyze the acoustic features of the speech based on The decoding network and the acoustic model are forced to align, and the result of segment segmentation is obtained by decoding. [0003] like figure 1 Shown is a flow chart of the traditional oral evaluation method, including the followin...

AI Technical Summary

Problems solved by technology

[0010] In the actual oral evaluation process of the above-mentioned traditional oral evaluation method, the following situations are likely to occur: the user does not read aloud according to the order of the text, the user reads content that has nothing to do with the text, or the environment is noisy. When these situations occur, it will lead to The segment segmentation boundary obtained is likely to be abnormal (or incorrect), and the subsequent evaluation process is based on the abnormal segment segmentation boundary, which will make the final output evaluation result unreliable and reduce the evaluation performance. The effect and user experience cannot meet the needs of users

Images