摘要
针对非特定文本的说话人识别,研究了特征提取方法及SVM核函数和参数选取对识别结果的影响,分析了现有的语音特征提取算法及各自的优缺点,以及不同核函数、核参数及惩罚参数对识别性能的影响.采用改进的网格寻优方法,进一步提高语音信息的识别时间.
The paper chose a fold that a non-specific text speaker identification. The paper focused on feature extraction methods and SVM kernel function and parameter selection on the identification re-suits, focusing on analysis of the existing voice feature extraction algorithms , their advantages and disadvantages, different kernel function, kernel parameters and penalty parameters on the recognition performance. Grid search method is introduced in order to improve the recognition time.
出处
《武汉理工大学学报(交通科学与工程版)》
2014年第2期316-319,共4页
Journal of Wuhan University of Technology(Transportation Science & Engineering)
基金
国家自然科学基金项目资助(批准号:51211130307)
关键词
支持向量机(SVM)
特征参数
核函数
识别模型
网格寻优算法
support vector machine (SVM)
feature Extraction
kernel function
recognition model
grid search method