基于人机交互的大词汇量连续语音自动识别系统设计  

Design of large vocabulary continuous speech automatic recognition system based on human-computer interaction

作　　者：曹琼茹 CAO Qiongru(Xinjiang University of Engineering,Urumqi Xinjiang 830000)

机构地区：[1]新疆工程学院,乌鲁木齐830000

出　　处：《自动化与仪器仪表》2021年第4期94-97,共4页Automation & Instrumentation

基　　金：中国语言资源保护工程专项任务:新疆汉语方言调查·奇台(No.YB1727A001)。

摘　　要：为了提高大词汇量连续语音自动识别能力,设计基于人机交互设计大词汇量连续语音自动识别系统。构建大词汇量连续语音信号检测模型,采用多尺度小波特征分解方法提取高阶频谱特征,并进行融合处理,构建特征多源辨识模型,通过空间波束集成方法检测大词汇量连续语音信号的信息聚类和语义特征,采用级联匹配滤波器进行大词汇量连续语音信号降噪滤波,在人机交互环境下设计系统硬件部分,至此完成基于人机交互的大词汇量连续语音自动识别系统设计。实验结果表明,采用该方法进行大词汇量连续语音自动识别的精度较高,特征匹配度较好,提高了大词汇量连续语音自动识别和检测能力。In order to improve the ability of automatic continuous speech recognition with large vocabulary,an automatic recognition system for continuous speech with large vocabulary is designed based on human-computer interaction.Construct a large vocabulary continuous speech signal detection model,use multi-scale wavelet feature decomposition to extract high-order spectral features,and perform fusion processing,build a feature multi-source identification model,and detect the information aggregation of large vocabulary continuous speech signals through spatial beam integration.For class and semantic features,cascaded matched filters are used for noise reduction and filtering of large vocabulary continuous speech signals.The hardware part of the system is designed in a human-computer interaction environment.So far,the design of a large vocabulary continuous speech automatic recognition system based on human-computer interaction is completed.The experimental results show that this method has higher accuracy and better feature matching for continuous speech recognition with large vocabulary,which improves the ability of automatic recognition and detection of continuous speech with large vocabulary.

关 键 词：人机交互 大词汇量 连续语音 自动识别 小波特征分解 信号检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]