噪声环境下应用于语音标注的端点检测算法研究
Endpoint detection algorithm for speech annotation in noisy environment
  
DOI:
中文关键词:  端点检测;信噪比分类;子带谱熵;子带频带方差;语音标注
英文关键词:endpoint detection;signal to noise ratio (SNR) classification;sub band spectral entropy;sub band frequency band variance;speech annotation
基金项目:
作者单位
俞景彦 同济大学 电子与信息工程学院,上海201804 
赵晓群 同济大学 电子与信息工程学院,上海201804 
摘要点击次数: 186
全文下载次数: 50
中文摘要:
      端点检测是语音标注的重要前序技术,针对语音标注,设计了一种基于信噪比分类的自适应端点检测算法。该算法首先对标注语音的信噪比分布范围进行分析,将信噪比分类,在每类信噪比范围内选择对应较优的算法。在高信噪比范围选择子带谱熵法,在中等信噪比范围内选择均匀子带频带方差法,而在低信噪比环境下先对带噪语音进行谱减法去噪处理,再采用基于均匀子带频带方差的端点检测算法。仿真实验表明,对语音标注采用的音频信号进行端点检测,该算法能达到较高的检测正确率,证明了算法的有效性。
英文摘要:
      The endpoint detection is an important preamble technology for speech annotation. For speech annotation, an adaptive endpoint detection algorithm based on signal to noise ratio (SNR) classification is designed. Firstly, the distribution range of the labeled speech SNR is analyzed and the SNR is classified by the algorithm. Then, the corresponding better algorithm is selected in the range of each type of the SNR. In the high signal to noise ratio range, the sub band spectral entropy method is selected; in the medium signal to noise ratio range, the uniform sub band frequency band variance method is selected. In a low signal to noise ratio environment, the noisy speech is firstly subjected to spectral subtraction denoising processing, and then an endpoint detection algorithm based on uniform sub band frequency band variance is used. Simulation experiments show that the endpoint detection algorithm for audio signals in the speech annotation can achieve a high detection accuracy rate, thus proving the effectiveness of the algorithm.
查看全文  查看/发表评论  下载PDF阅读器

你是第2142801访问者
版权所有《南京邮电大学学报(自然科学版)》编辑部
Tel:86-25-85866913 E-mail:xb@njupt.edu.cn
技术支持:本系统由北京勤云科技发展有限公司设计

欢迎访问《南京邮电大学学报(自然科学版)》编辑部!