基于错误信息提取的考场语音识别纠错模块设计

Design of Error Correction Module for Speech Recognition in Examination Room Based on Error Information Extraction

为了解决现有语音识别系统对错误发音识别精度不高的问题,提出并设计了一种发音自动纠错技术。分别通过CCD传感器和TDSP-TF887芯片对系统硬件部分进行了优化设计,软件部分引入DTW算法实现了考场专用错误语音的智能识别。系统测试结果表明,本研究所设计的系统对错误语音的识别准确率高出其他同类系统36.5%,具有很强的实用性。

In order to solve the problem that the existing speech recognition system has low recognition accuracy for wrong pronunciation,an automatic pronunciation correction technology is proposed and designed.The hardware part of the system is optimized through the CCD sensor and TDSP-TF887 chip respectively,and the DTW algorithm is introduced into the software part to realize the intelligent recognition of the special error speech in the examination room.The system test results show that the recognition accuracy of the system designed in this study is 36.5%higher than other similar systems,and has strong practicability.