机载多通道语音信号自适应合成

Adaptive Synthesis of Airborne Multi-channel Speech Signals

机载环境下,超短波通信系统由于电磁干扰等诸多原因,容易导致接收到的语音出现断续。引入基于空-时模型的线性约束最小方差(Spatio-Temporal Linearly Constrained Minimum Variance,ST-LCMV)方法来对多通道的语音信号进行合成,保证话音的完整性,增强话音的稳定性。构建了多通道语音信号合成(Multi-channel Speech Signal Synthesis,MSSS)算法并使用真实的机载噪声进行仿真验证,结果表明,MSSS能保留完整的语音信息,并且在实验设置的机载噪声环境下具有普适性。MSSS输出信号的语音感知质量评估(Perceptual Evaluation of Speech Quality,PESQ)分数相对于等增益合并提高了0.4518分。

In airborne environment,the UHF/VHF communication system is easy to cause the received speech signals to be intermittent due to many reasons,such as electromagnetic interference and so on.To solve this problem,spatio-temporal linearly constrained minimum variance(ST-LCMV)is introduced to synthesize multi-channel speech signal and ensure the integrity of speech and enhance the stability of speech.Multi-channel speech signal synthesis(MSSS)algorithm is constructed and verified by simulation with real airborne noise.Through the simulations,it can be known that MSSS can retain complete speech information and is universal in airborne noise environment of the experiments.The perceptual evaluation of speech quality(PESQ)score of MSSS’s output signals is increased by 0.4518 points compared with that of Equal Gain Combining(EGC)on average.