摘要
为了提高声带疾病嗓音识别率,提出了一种基于差异度的多参数识别方法。研究声学参数之间的相关性,定义两类嗓音间参数的差异度;根据差异度对参数进行权重分配,并剔除相关性较大的参数,然后对参数加权组合为新的特征向量。以DSP芯片TMS320VC5502为核心实现识别系统,识别实验表明,新的参数在声带疾病嗓音与正常嗓音识别中识别率有4.35%的提升,在声带疾病嗓音与非声带疾病嗓音识别中有9.27%的识别率提升。
A method based on recognition rate. It studies the multi-parameter with degree of difference is proposed to correlations between acoustic parameters and the degree improve vocal of differences cord disease between two kinds of voice. According to the degree of difference, the acoustic parameters are weight distributed and the parameters with large correlation are removed. The new parameters are obtained after weighting and recombining. It takes TMS320VC5502 as center of recognition system. The experimental results show that the using of new parameters gets the 4.35% promotion in recognition of vocal cord disease voice and normal voice, as well as 9.27% recognition rate increase in the identification of vocal cord disease voice and non-vocal cord diseases voice.
出处
《电子器件》
CAS
北大核心
2013年第3期309-312,共4页
Chinese Journal of Electron Devices
基金
国家自然科学基金项目(61271359
61071215)
苏州大学捷美生物医学工程仪器联合实验室资助项目
关键词
声带疾病嗓音识别
差异度
声学参数
DSP
vocal cord disease voice recognition
degree of difference
acoustic parameters
DSP
