Description: 说话人识别中,最后带测语音匹配阶段.利用DTW算法实现-Speaker recognition, the last swath of voice matching stage. The use of DTW algorithm Platform: |
Size: 1024 |
Author:雷硕 |
Hits:
Description: matlab语音信号处理工具箱
本工具箱可配合语音数据库使用，用于计算线性预测语音模型的参数，语音声调转换，语音自动解析分解，语音语速变换及更改发音的重音，音量，清晰度等。配套图书《Speech Processing and Synthesis Toolboxes》已经出版，翻译进行中-matlab signal processing toolbox with the voice of the Toolbox database can be used for calculating linear predictive speech model parameters, voice tone conversion, automatic speech analysis decomposition, transformation and change of voice speech rate accented pronunciation, volume, clarity, etc. . Matching book " Speech Processing and Synthesis Toolboxes" has been published in translations Platform: |
Size: 9350144 |
Author:wujingjing |
Hits:
Description: VQ声纹识别算法和实验.
摘要：采用线性预测倒谱系数(1inear prediction cepstrum coefficient，LPCC)作为语音的特征参数，矢量量化(vector quantity，VQ)方法进行模式匹配，探讨声纹识别以实现身份认证，并对此识别方法进行了相关的实验．通过验证，这种方法可以区分不同的说话人，并且在做说话人辨认实验时可达到较高的识别率．-VQ pattern recognition algorithms and experimental sound. Abstract: The linear prediction cepstral coefficients (1inear prediction cepstrum coefficient, LPCC) as the voice of the characteristic parameters, vector quantization (vector quantity, VQ) method of pattern matching, pattern recognition of sound in order to achieve authentication and identification methods relevant to this experiment. Validated, this method can distinguish between different speaker, and speaker identification experiments to do to reach a higher recognition rate. Platform: |
Size: 171008 |
Author:海边贝壳 |
Hits:
Description: 于市售 STM32 开发板上实现特定人语音识别处理项目。识别流程是：预滤波、ADC、分帧、端点检测、预加重、加窗、特征提取、特征匹配。端点检测(VAD)采用短时幅度和短时过零率相结合。检测出有效语音后，根据人耳听觉感知特性,计算每帧语音的 Mel 频率倒谱系数(MFCC)。然后采用动态时间弯折(DTW)算法与特征模板相匹配,最终输出识别结果。先用Matlab对上述算法进行仿真，经数次试验求得算法内所需各系数的最优值。而后将算法移植到 STM32 开发板上，移植过程中根据 STM32 上存储空间相对较小、计算能力也相对较弱的实际情况，对算法进行优化。最终完成于 STM32 微处理器上的特定人语音识别系统。-Implement speech recognition processing project in commercially available STM32 development board. Identification is the process: pre-filter, ADC, framing, endpoint detection, pre-emphasis, windowing, feature extraction, feature matching. Endpoint detection (VAD) short-time amplitude and short-term zero rate combined. After detecting an effective voice, according to the characteristics of human auditory perception, calculated for each frame of speech Mel Frequency Cepstral Coefficients (MFCC). Then dynamic time warping (DTW) algorithm and feature template matches the final output recognition result. First with Matlab simulation algorithm described above, after several trials to get the optimal value of each coefficient within the desired algorithm. The algorithm will migrate to STM32 development board, the porting process according to the STM32 relatively small storage space, computing power is relatively weak situation of the optimization algorithm. Finally completed on the STM32 micr Platform: |
Size: 325632 |
Author:Chenkly |
Hits:
Search - voice matching in matlab