语音信号处理pdf 赵力

谱减法的程序 希望可以被大家使用～～ 做的比较一般但运行起来还是可以的(Spectral subtraction procedure can be hoped that the use of ~ ~ U.S. to do the more general but can be up and running the)

简单语音识别源码是一个调用讯飞语音识别SDK的例子源码是一个最纯净的Demo比较容易看懂。实现的是点击按钮开始语音监听，手机需要联网，2/3G的均可，识别后会以Toast的方式提示出来，识别率还是比较高的。之前自己是用别人的Demo总是不成功，是因为，讯飞的SDK想使用是需要APPID的，可能过太长时间久会注销，还有，未通过审核的APPID每天只能试用500次，如果太多人试用，就是达到上限导致无法使用。(A simple speech recognition source code is a call Ifly speech recognition SDK example source code is a demo of the purest easier to understand. The realization of the click button to start the voice listening, mobile phone need to networking, 2/3G can, after the recognition will be prompted by Toast way out, the recognition rate is still relatively high. Before he is with someone else s demo is not always successful, is because the Xunfei SDK to use is appid, may be too long for a long time will be canceled, and not through the audit appid every day can only be tried out 500 times, if too many people tried is reaching the limit to use.)

1、文件夹中包含了经典功率谱估计和自适应均衡算法两个实验的所有程序。 2、R.m、LMS.m、LMSmain.m为自适应均衡算法的程序： R.m用来计算输入信号的自相关矩阵及其特征值； LMS.m为时域LMS算法，用统计的方法仿真得出不同信道参数和不同步长下的学习曲线； LMSmain.m为实验主程序，按照实验要求中的具体数据得到实验结果和曲线。 3、functionx.m、fzhouqitu.m、spectrum.m、bt.m、bart_lett.m、welch.m、SPECTRUMmain.m为经典谱估计的程序： functionx.m产生需要进行谱估计的函数； fzhouqitu.m用来计算信号周期图的函数； spectrum.m是用周期图法进行谱估计的函数 bt.m是用BT图法进行谱估计的函数 bart_lett.m是用BARTLETT法进行谱估计的函数 welch.m是用WELCH法进行谱估计的函数 SPECTRUMmain.m是主程序，按照实验要求中的具体数据得到实验结果和曲线。 (a document folder contains the classic power spectrum estimation and adaptive equalization algorithm for the two experiments all the procedures. 2, R.m, LMS.m, LMSmain.m adaptive equalization algorithm for the procedure : R.m used to calculate the input signal and the correlation matrix eigenvalue; LMS.m too LMS algorithm, use statistical simulation method come to a different channel parameters and the synchronous learning curve; LMSmain.m main program for the experiment, according to the experimental requirements of the specific data and the curve of the experimental results. 3, functionx.m, fzhouqitu.m, spectrum.m. bt.m, bart_lett.m, welch.m. SPECTRUMmain.m classical spectrum estimation procedures : functionx.m have a need for spectrum estimation function; fzhouqitu.m used to calculate t)

语音信号处理，阵列为10阵元线阵，宽带和窄带波束形成(Speech signal processing, 10 sensors of microphone linear array,Broadband and narrowband beamforming)

characteristic parameters for speech recognition on the extraction function

房间脉冲响应函数IMAGE,用于计算房间混响时间的(Room impulse response function IMAGE, used to calculate the room reverberation time)

多抽样率信号处理，包括邓氏关联度、绝对关联度、斜率关联度、改进绝对关联度,完整的基于HMM的语音识别系统。( Multirate signal processing, Including Deng s correlation, absolute correlation, correlation of slope, improved absolute correlation, Complete HMM-based speech recognition system.)

短时自相关和短时平均幅度差计算基音周期的matlab程序，附带有原语音信号(Short-term autocorrelation and short-term average deviation of pitch period matlab program, comes with the original speech signal)

利用自适应滤波器对心音信号进行预测，包括了LMS算法，归一化LMS，变步长LMS和RLS算法，其中RLS的预测效果最好.(Heart sound signals using adaptive filter to predict, including the LMS algorithm, the normalized LMS, variable step size LMS and RLS algorithm, in which the best forecast of RLS.)

说明：  基于CRF的中文机构名识别系统。使用北京大学1998年的人民日报语料库作为训练语料。除常用的特征模板，已经词性特征外，使用词语的最后一个字作为特征，提高了机构名识别的准确率, 调用了CRF++程序包训练模型。(CRF-based name recognition system of Chinese institutions. People' s Daily, Peking University in 1998 with corpus as training data. In addition to the characteristics of commonly used templates, has been part of speech features, the last word to use words as features and improve the organization name recognition accuracy, call the CRF++ package training model.)