欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
提出了一种将改进的FCM聚类算法与矢量量化相结合的说话人识别的方法。先从语音信号中提取待识别的特征矢量集，再利用矢量量化来设计码本，最后用改进的算法对待识别语音进行辩识。该算法解决了FCM算法对初始值敏感、易陷入局部最优的问题。所使用的特征参数较少，计算比较简单，但识别率较高，且具有较好的鲁棒性。
英文摘要：
A speaker recognition method based on improved FCM clustering with vector quantization is introduced in this paper. Firstly, feature vectors are extracted from speech signal, secondly, vector quantization is utilized to design code book,and lastly,improved method is used to recognize the speech signal, The FCM algorithm is quite sensitive to initial value and tends to local optimization, while the method proposed in this paper has solved these problems with less feature parameters and computation, the experiments result reveals it has higher recognition rate and better robustness.
同期刊论文项目
同项目期刊论文
期刊信息
位置：