文章基本信息

标题：Performance Analysis of Learning Classifiers for Spoken Digit Under Noisy Conditions
本地全文：下载
作者：Syed Abbas Ali ; Samreen Burkiand ; Sara Hasan 等
期刊名称：Journal of Emerging Trends in Computing and Information Sciences
电子版ISSN：2079-8407
出版年度：2013
卷号：4
期号：3
页码：286-289
出版社：ARPN Publishers
摘要：Speech Recognition is truly challenging task due to the presence of ambient noise. Ambient noise can decrease the Intelligibility and reliability of the speech recognition system which causes misinterpretation of speech sound. Most of the practical applications of speech processing, like audio conferencing, cellular communication and the recorded speech contain a significant amount of ambient noise. In this paper, we evaluate the performance of different classifiers (such as J48, Naïve Bayes, Multiclass and Multilayer Perceptron) in term of classifiers recognition rate for recorded isolated digit and TIDIGIT corpus. The noise and speech samples taken from NOISEX-92 database and TIDIGIT corpus respectively, and recorded isolated digits (0-9) taken from real environment. The experimental results clearly show that the Multilayer Perceptron and Naïve Bayes classifiers indicate maximum recognition rate in comparison with Multiclass classifier, while J48 classifier shows minimum recognition rate for recorded isolated digits and TIDIGIT corpus.
关键词：Learning Classifier; Recognition Rate; Ambient noise; WEKA Data Mining; Hidden Markov Model (HMM)