Abstract
In this paper, a new method for making v/uv decision is developed which uses a multi-feature v/uv classification algorithm based on the analysis of cepstral peak, zero crossing rate, and autocorrelation function (ACF) peak of short-time segments of the speech signal by using some clustering methods. This v/uv classifier achieved excellent results for identification of voiced and unvoiced segments of speech.
Highlights
A new method for making v/uv decision is developed which uses a multi-feature v/uv classification algorithm based on the analysis of cepstral peak, zero crossing rate, and autocorrelation function (ACF) peak of short-time segments of the speech signal by using some clustering methods
The voiced/unvoieced decision is critical in many speech analysis/synthesis systems because it is essential to know whether the speech production system involves vibration of the vocal cords [1,2,3,4]
The V/UV decision is made in this way: each of the information groups is clustered into three clusters by K-Means algorithm [23]
Summary
The voiced/unvoieced decision is critical in many speech analysis/synthesis systems because it is essential to know whether the speech production system involves vibration of the vocal cords [1,2,3,4] This decision is required for many applications, including modeling for analysis/synthesis, detection of model changes for segmentation purposes and signal characterization for indexing and recognition applications [1]. Common methods extract a feature from speech segments and make the v/uv decision according to whether the value of the feature is above or below a pre-determined threshold. The use of two or more features in the voicing decision tended to the methods which do not consider the features as a vector and use the best feature for each frame like the work in [20].
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
