a little app to turn text to speech

AMR-WB声码器VAD（语音活动检测）程序。测试使用8kHz采样率和16位wav量化文件。

赤木正人教授在论文“Development of an F0 control model based on F0 dynamic characteristics for singing-voice synthesis”中提出的f0 vibrato模型的matlab实现-Professor Masato Akagi paper "Development of an F0 control model based on F0 dynamic characteristics for singing-voice synthesis" presented f0 vibrato model matlab implementation

一个mp3的解码程序，c语言实现。其中只对layer III 进行解码。-A program of mp3 decode,programmed by c language.It can only decode layer 3.

这是一个通用的隐马尔科夫模型的C代码库， 非常好用。-This is a generic Hidden Markov Model C code libraries, very handy.

数字音频格式有很多种，其质量与采样频率和采样精度两个参数有关。频率的表示单位为赫兹〔Hz〕，它表示每秒采样次数。采样频率越高，音质就越好。采样精度为每次采样所存储的数据数量，它决定每个数字信号所能够表示的离散振幅的数量。存储每个样本的数据越多，音质就越好。但是高品质的声音需要占用大量的内存和磁盘空间。考虑到网络带宽，在Internet连接上传输就需要花费很长的时间。对于Applet来说，保证声音文件的最小化是极为重要的。-digital audio format there are many, the quality and frequency of sampling and sampling accuracy of the two parameters. The frequency of Hertz unit of [Hz], it said the number of samples per second. The higher sampling frequency, the sound quality better. For each sampling precision sampling by the amount of data storage, it was decided that each digital signal can be expressed as the number of discrete amplitude. Each sample storage of more data, better sound quality. But high-quality voice would take a lot of memory and disk space. Consideration of the network bandwidth, the transmission on the Internet connection on the need to spend a very long time. Applet for instance, to ensure that voices of the smallest document is very important.

对于语音和图像的数字信号处理基本知识的介绍，对信号处理的各个基本知识可以有个大概的认识-A rough introduction for the dsp of speech and video.It could give you a basic knowledge of signal processing.

:高斯混合模型(GMM)是一种经典的说话人识别算法，本文在实现其算法的同时，主要模拟了不同噪声环境情况下高斯混合模型 (GMM)的杭嗓声性能，得到了一些有益结论。 -Gaussian mixture model (GMM) is a classic speaker recognition algorithms, this algorithm at the same time in fulfilling its main simulated environmental conditions under different noise Gaussian mixture model (GMM) of the Hang throat sound performance, and obtained some useful conclusions.

LMS toolbox si a matlab code for least mean square algerion-LMS si a toolbox for Matlab code"s least mean quare algerion

The Yin algorithm was developed by Alain de Cheveigné of IRCAM-CNRS and Hideki Kawahara of Wakayama University. It allows for real-time continuous (for each sample) fundamental frequency estimation. It features a very low error rate and few tuning parameters. This implementation is a C++ template, with EMM support for faster processing. Developped in Visual C++ 6.0, but should compile with other compilers.-The Yin algorithm was developed by Alain de Cheveigné of IRCAM-CNRS and Hideki Kawahara of Wakayama University. It allows for real-time continuous (for each sample) fundamental frequency estimation. It features a very low error rate and few tuning parameters. This implementation is a C++ template, with EMM support for faster processing. Developped in Visual C++ 6.0, but should compile with other compilers.

用于语音识别的汉语树形词典，完成了6万词汇的建树操作及相关子节点等的查询处理，主要用于指导语音搜索整个建立时间大概为3分钟-for speech recognition dictionary of Chinese tree, completed 60,000 contribution to the vocabulary and related operations, such as sub-nodes of query processing, Pronunciation guide for the main search the entire building for about three hours minutes