融合改进GCS算法的语音增强技术对同声传译设备的优化研究  

Research on the optimization of simultaneous interpretation equipment based on voice enhancement technology integrated with improved GCS algorithm

作　　者：冯掬琳[1] 方亚利[1] FENG Julin;Fang Yali(Yulin University,Yulin Shaanxi 719000,China)

机构地区：[1]榆林学院,陕西榆林719000

出　　处：《自动化与仪器仪表》2023年第8期268-272,共5页Automation & Instrumentation

摘　　要：针对同声传译过程中的语音特征识别失真问题,提出了一种以广义旁瓣消除(Generalized Sidelobe Cancelling,GCS)算法架构为基础,融合自适应滤波器的改进式语音增强模型,该模型通过控制信号的相干性和能量比来实现自适应噪声相消器(Adaptive Noise Canceller,ANC)模块的更新。同时研究为避免期望信号在ANC模块的对消现象,通过联合判定因子来控制ANC模块的更新。研究结果显示,随着信噪比输入的增加,研究设计的模型的客观语音质量评估(Perceptual Evaluation Of Speech Quality,PESQ)数值呈现出前期迅速,后期缓慢的增长趋势,增长区间为1.8到2.6。同时在四种不同的噪声干扰环境下,研究设计模型的PESQ均值、频域分段信噪比均值均大于对比模型。不同干扰角度与期望信号距离下模型的PESQ均值同样都更高。由此可见研究设计的模型更具性能优势,更有能力保证语音信息不失真。Aiming at the problem of speech feature recognition distortion in the process of simultaneous interpretation,an improved Speech enhancement model based on generalized sidelobe cancellation(GCS)algorithm architecture and fusion of adaptive filters is proposed.This model updates the adaptive noise canceller(ANC)module by controlling the signal coherence and energy ratio.At the same time,in order to avoid the cancellation phenomenon of expected signals in the ANC module,joint decision factors are used to control the update of the ANC module.The research results show that as the signal-to-noise ratio input increases,the perceptual evaluation of speech quality(PESQ)values of the model designed in the study show a rapid growth trend in the early stage and a slow growth trend in the later stage,with a growth range of 1.8 to 2.6.At the same time,under four different noise interference environments,the PESQ mean and frequency-domain segmented signal-to-noise ratio mean of the research and design model were greater than those of the comparison model.The PESQ mean of the model is also higher under different interference angles and expected signal distances.From this,it can be seen that the model designed through research has more performance advantages and the ability to ensure that speech information is not distorted.

关 键 词：GCS 语音增强 同声传译 阵列 噪声 

分 类 号：TN912.35[电子电信—通信与信息系统]