The Steered Response Power(SRP)method works well for sound source localization in noisy and reverberant environment.However,the large computation complexity limits its practical application.In this paper,a fast SRP search method is proposed to reduce the computational complexity using small-aperture microphone array.The proposed method inspired by the SRP spatial spectrum includes two steps:first,the proposed method estimates the azimuth of the sound source roughly and determines whether the sound source is in far field or near field;then,different fine searching operations are performed according to the sound source being in far field or near field.Experiments both in simulation environments and real environments have been performed to compare the localization accuracy and computation complexity of the proposed method with those of the conventional SRP-PHAT algorithm.The results show that,the proposed method has a comparative accuracy with the conventional SRP algorithm,and achieves a reduction of 93.62%in computation complexity compared to the conventional SRP algorithm.
针对传统的SRP-PHAT(Steered Response Power with Phase Transform)声源定位算法容易受噪声影响而导致定位性能降低的问题,提出一种频域补零且保留部分镜像分量的改进算法。该算法首先通过傅里叶变换将接收信号变换到频域,然后在高频端补零至20倍帧长,同时保留部分镜像分量。在此基础上计算麦克风对接收信号的互功率谱密度函数,作傅里叶逆变换得到相位变换加权的广义互相关(GCC-PHAT)函数。保留的镜像分量拓宽了信号频域,使GCC-PHAT函数的峰更为尖锐,累加后得到的SRPPHAT函数的空间谱峰也就更加尖锐,从而提高定位性能。实验表明,相比于传统算法,改进算法能显著提高定位成功率。
To improve the recognition performance of video human actions,an approach that models the video actions in a hierarchical way is proposed. This hierarchical model summarizes the action contents with different spatio-temporal domains according to the properties of human body movement.First,the temporal gradient combined with the constraint of coherent motion pattern is utilized to extract stable and dense motion features that are viewed as point features,then the mean-shift clustering algorithm with the adaptive scale kernel is used to label these features.After pooling the features with the same label to generate part-based representation,the visual word responses within one large scale volume are collected as video object representation.On the benchmark KTH（Kungliga Tekniska H？gskolan）and UCF （University of Central Florida）-sports action datasets,the experimental results show that the proposed method enhances the representative and discriminative power of action features, and improves recognition rates.Compared with other related literature,the proposed method obtains superior performance.
将矢量泰勒级数(Vector Taylor Series,VTS)特征补偿算法应用于说话人识别,给出了卷积噪声方差的近似闭式解,构建了联合快速估计卷积噪声和加性噪声均值和方差的框架。该算法可在无需失配环境先验信息的前提下,直接从失配语音中估计出卷积噪声和加性噪声的均值和方差,实现对环境失配的补偿。实验结果表明,在信道变化较大的无线信道下,卷积噪声方差的补偿最高可降低误识率3.24%.提升了系统的识别性能。在存在加性噪声的无线信道下,与基于线性失真模型的特征映射算法和倒谱均值减算法相比,本文算法可分别最大降低49.65%和68.06%的误识率,适合于信道变化较大的失配环境补偿。
