基于声场感知的多声源自动增益控制技术  

Multi-source Automatic Gain Control Based on Sound Field Perception

作　　者：卢佳欣 陈枢茜 朱阳燕 LU Jiaxin;CHEN Shuxi;ZHU Yangyan(Nantong Institute of Technology,Nantong 226001,China)

机构地区：[1]南通理工学院,江苏南通226001

出　　处：《电声技术》2023年第6期47-51,共5页Audio Engineering

基　　金：江苏省高校基础科学(自然科学)研究基金面上资助经费项目(22KJB520032);教育部供需对接就业育人项目(20230109397)。

摘　　要：传统的单通道自动增益控制技术在会议室等多人说话音频通信场景下存在无法自适应跟踪多个说话人的响度水平、音频音量响度在时间上不连续以及频率失真等问题,文章提出一种基于声场感知的多声源自动增益控制技术。利用声场感知算法估计声源时空相似度向量,根据该向量自适应控制声源响度的追踪速度,从而提升声源能量的时域追踪准确率。应用基于混响强度估计的自适应增益补偿技术,改善声音高低频成分混响差异过大和能量衰减不同的问题。实验结果表明,在声源切换情况下,所提出的基于声场感知的多声源自动增益控制技术有效提升了声源音量响度一致性,音频信号的平均能量为-5.22 dB,能量波动区间为[-8.76,-3.46] dB,能量位于该区间的概率为99.3%。Because the traditional single-channel automatic gain control technology can't adaptively track the loudness level of multiple speakers,the loudness of audio volume is discontinuous in time,and the frequency is distorted,etc.,this paper proposes a multi-sound source automatic gain control technology based on sound field perception.The space-time similarity vector of sound source is estimated by sound field perception algorithm,and the tracking speed of sound source loudness is adaptively controlled according to this vector,thus improving the accuracy of time-domain tracking of sound source energy.The adaptive gain compensation technology based on reverberation intensity estimation is applied to solve the problems of excessive reverberation difference and different energy attenuation of high and low frequency components of sound.The experimental results show that,in the case of sound source switching,the proposed multi-source automatic gain control technology based on sound field perception effectively improves the consistency of sound source volume and loudness.The average energy of audio signal is-5.22 dB,the energy fluctuation range is[-8.76,-3.46]dB,and the probability that the energy lies in this range is 99.3%.

关 键 词：自动增益控制 最大似然估计 声源定位 听觉滤波器组 

分 类 号：TN641[电子电信—电路与系统]