矿井无线透地通信语音压缩与识别仿真

摘要

矿井无线透地通信系统可以实现地面与井下的实时或低时延语音通信,其中语音信号的正确识别能够保证通信的有效性.针对无线透地通信中信道资源十分紧张的问题,为使通信系统在工作时占用较少的信道带宽,提出了将混合激励线性预测(Mixed-Excitation Linear Prediction,MELP)编码方案应用于矿井无线透地通信中对语音信号进行压缩处理,并提出进一步降低编码速率的类MELP方案.利用Matlab对上述方案进行仿真验证,将压缩后的语音信号通过无线透地通信仿真信道并重构语音,经实际测听,恢复后的语音信号能够被人为主观识别,其MOS(Mean Opinion Score)评分接近4.8kb/s的CELP编码.仿真结果表明,上述方案基本满足无线透地通信对语音压缩与识别的要求.%The mine wireless communication system can realize the real-time or low-delay voice communication between the ground and the underground,in which the correct recognition of the voice signal can guarantee the validity of the communication.The channel resources of wireless through-the-earth communication are very limited.In order to make the communication system occupy less channel bandwidth at work,we proposed a method to apply the Mixed-Excitation Linear Prediction MELP coding scheme to mine wireless through-the-earth communication for speech compressing.On this basis,we further reduced the coding rate of the class MELP program.Using Matlab to simulate the scheme,the compressed speech signals transited through the simulation channel of wireless through-the -earth communication and was reconstructed.The reconstructed voice signals can be subjectively identified by the actual audiometry,and the Mean Opinion Score (MOS) score is close to 4.8kb/s CELP's.The simulation results show that the above schemes basically meet the requirements of voice compression and recognition for wireless through-the -earth communication.