基于BLSTM网络的智能客服语音识别系统研究

Research on Intelligent Customer Service Speech Recognition System Based on BLSTM Network

为了提高智能客服服务质量,提出基于BLSTM网络的智能客服语音识别系统。建立了系统总体构架,设计了系统硬件及软件。通过PCI9054进行语音识别系统的集成DSP控制,采用低功耗的GT834032进行智能客服语音识别的集成语音信息处理和指令加载,完成硬件设计。软件部分根据匹配滤波检测方法完成信号增强和滤波处理,在时域和频域空间内提取高阶谱特征量,实现波束集成处理。通过网络信道均衡调制方法均衡控制语音传输,结合稳定性测试方法完成稳定特征提取和定位检测,实现智能客服语音信号识别。仿真实验结果表明,所设计系统对语音信号识别的抗干扰性较好,特征辨识度较高,信号检测和识别能力较强。

To improve the service quality of intelligent customer service,an intelligent customer service speech recognition system based on BLSTM network is proposed.We construct the overall system framework,and design the system hardware and software.PCI9054 is used to control the integrated DSP of speech recognition system,and low-power GT834032 is used for integrated speech information processing and instruction loading of intelligent customer service speech recognition,and the hardware design is completed.According to the matched filter detection method,the software completes signal enhancement and filtering processing,and extracts high-order spectral features in time domain and frequency domain space to realize beam integration processing.Balanced control of voice transmission is used by network channel equalization modulation method.Combined with stability test method stable feature extraction and location detection are completed,intelligent customer service voice signal recognition is realized.The simulation results show that the designed system has good anti-interference performance,high feature recognition degree,strong signal detection and recognition ability.