基于声学模型的不良语音识别技术研究

不良语音识别技术是在高清音视频业务中监测不良信息的有效手段。本文提出了一个基于声学模型的不良语音识别框架。并针对框架中特征提取、声学模型构建、不良语音判定模型三个重要部分的实现方法进行分析描述。并列举了各种方法的优缺点。对构建高效的不良语音识别系统具有较高的参考价值。

Research on Bad Speech Recognition Technology Based on Acoustic Model

Bad speech recognition technology is an effective means to monitor bad information in high-definition audio and video services. This paper presents a framework for bad speech recognition based on acoustic models. The paper analyzes and describes the implementation methods of three important parts: feature extraction, acoustic model construction and bad speech judgment model. And enumerate the advantages and disadvantages of various methods. It has a high reference value for constructing an efficient bad speech recognition system.