摘要
数字摄像机的摄像头在缩放时产生较大声噪,回放时主观听觉十分明显,因此需要研究一种高质量语音降噪算法,用于消除机械声噪,同时避免产生语音畸变。通过分析传声器阵列及带噪声源的数学模型,构建了一套基于传声器阵列的语音降噪系统,针对机械噪声的统计特性,利用最小方差无畸变响应波束形成法对非注视方向的噪声进行滤波,采用子带后处理法进一步降噪。该算法在保证稳定性的前提下降低了语音降噪处理的复杂度并提高信噪比;提出一种新的单通道维纳滤波函数,降低了信号低频部分的畸变。实验语音数据均在普通会议室采集,新的增强算法增益为16dB,主观听觉可懂度较好。实验结果证明,该系统可作为理想的语音降噪设备应用于摄像机录制过程中。
Digital video camcorder is a device that records/stores images. But customers always hear a loud noise which is clearly picked up when the camera zooming in and out quickly during playback of video. Therefore a high-quality mechanical noise reduction algorithm which avoided speech distortion is needed to research. In this paper, the mathematic model of microphone array and noisy sound source is analyzed and one noise reduction system based on microphone array is built. By the statistic character of mechanical noise, MVDR Beamforming is utilized to filter the non-directed noise and subband postfiher is utilized to reduce more noise. The proposed approach is to reduce complexity and improve its accuracy under the premise of ensuring the stability. A new single channel wiener filter function is proposed to depress the distortion of signal low-frequency portion. Experiment＇s data is collected in the normal meeting room. The gain of enhancement algorithm is 16 dB and subjective impression is better. The results show that this system can be an ideal noise reduction device applied for camera recording process.
出处
《电视技术》
北大核心
2011年第9期86-89,共4页
Video Engineering
基金
国家自然科学基金委员会与微软亚洲研究院联合资助项目(60776800)
国家自然科学基金委员会与香港研究资助局联合科研基金资助项目(60931160443)
国家高技术研究发展计划(863计划)项目(2008AA040201
2008AA02Z414)
国家自然科学基金项目(90920302)
关键词
语音信号处理
数字摄像机
机械噪声
噪声消除
传声器阵列
speech signal processing
digital video camcorder
mechanical noise
noise reduction
microphone array
