文章基本信息

标题：Nonlinear Features of Bark Wavelet Sub-bandFiltering for Pathological Voice Recognition
本地全文：下载
作者：Xiao-Jun Zhang ; Xin-Cheng Zhu ; Di Wu 等
期刊名称：Engineering Letters
印刷版ISSN：1816-093X
电子版ISSN：1816-0948
出版年度：2021
卷号：29
期号：1
页码：49-60
语种：English
出版社：Newswood Ltd
摘要：This paper proposes a recognition method forextracting nonlinear characteristic parameters of pathologicalvoices using Bark wavelet sub-band filtering. First, the speechsignal was processed through 24 Bark filter banks. According tothe signal obtained from each channel, the featuremulti-frequency band nonlinear coefficient was extractedaccording to the frequency division factor, . We used 53normal voices and 117 pathological voices from the MEEI’spathological voice experimental database, and 14 machinelearning methods were used to perform the recognitionexperiments. The experimental results showed that theproposed method effectively improved the recognition rate.Moreover, the proposed feature was optimal when = 18, andthe highest recognition rate was achieved when the supportvector machine learning algorithm was used.
关键词：frequency division; nonlinear; pathological voice speech recognition; frequency division factor