基于深度学习的闽南语语音识别  

Speech Recognition of Minnan Dialect Based on Deep Learning

作　　者：郭新华 GUO Xinhua(Software Institute,Quanzhou University of Information Engineering,Quanzhou 36200,China)

机构地区：[1]泉州信息工程学院软件学院,福建泉州362000

出　　处：《智能物联技术》2023年第4期13-19,共7页Technology of Io T& AI

基　　金：2021年度福建省中青年教师教育科研项目(科技类)“基于深度学习的闽南方言语音情感识别”(JAT210552)。

摘　　要：目前的智能家居语音控制仍以普通话为主,较难进行方言的识别。闽南语作为福建第一大方言,使用人数众多,将闽南语语音识别用于智能家居系统将有广阔前景。为了提高闽南语语音识别模型的性能,本文将cFSMN声学模型、DFSMN声学模型和MTL-DFSMN声学模型分别在闽南语语音语料库上进行验证,最终选择声学模型MTL-DFSMN用于闽南语语音识别。该模型能更精准识别用户的闽南语语音指令和满足智能家居系统的要求。At present,the voice control of smart homes is still mainly in Mandarin,making it difficult to recognize di-alects.As the largest dialect in Fujian,Minnan dialect is used by a large number of people,and the application of Minnan dialect'speech recognition in smart homes will have broad prospects.In order to improve the performance of Minnan di-alect's speech recognition's model,the cFSMN model,DFSMN model and MTL-DFSMN model were respectively verified on Minnan dialect's speech recognition,so as to select a better acoustic model for the Minnan dialect's recognition.The re-sults showed that the MTL-DFSMN Acoustic model could recognize the user's Minnan dialect's voice commands more accu-rately and meet the requirements of smart home system.

关 键 词：深度学习 闽南方言语音识别 智能家居 多任务学习 深层前馈序列记忆神经网络 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]