Robust Feature Sets for the Speech Processing
Journal Title: Acta Technica Napocensis- Electronica-Telecomunicatii (Electronics and Telecommunications) - Year 2012, Vol 53, Issue 2
Abstract
This paper compares the various feature sets for speech signal such as cepstral-based and acoustically-driven parameters to analyze and see how robust they are in different noise environments. Despite the success of the cepstral-based features in the tasks of speech processing, they are still susceptible to noise. In this work, we explored how to extend MFCC-based feature sets with other acoustically-driven parameters to add some robustness for the segmentation of speech. Experiments conducted on TIMIT dataset using the standard HMM/GMM framework show that better performance can be achieved if cepstral-based features are combined with acoustic parameters.
Authors and Affiliations
Zhandos YESSENBAYEV
Keywords
Related Articles
- EP ID EP114287
- DOI -
- Views 115
- Downloads 0
Pairing For Elliptic Curves With Embedding Degree 1 Defined Over Prime Order Fields
In the last years, prime order pairing-based cryptography has received much attention. The embedding degree 1 is very appealing because the computations can be performed without any extension fields, but the efficiency p...
Human Visual Perception Concepts As Mechanisms For Saliency Detection
Predicting human eye fixations is a difficult task that usually involves complex knowledge about the human visual perception pathways. This paper proposes a method to detect salient locations using a psycho visual space...
Forecasting By Neural Networks In The Wavelet Domain
This paper presents a forecasting method for time series. This method combines the wavelet analysis and several forecasting techniques such as Artificial Neural Networks (ANN), linear regression and random walk. The prop...
Relevant Context Discovery for Pervasive Services Involving User Control
Dynamic context adaptation is a central concept in Pervasive and Ubiquitous Computing. In this paper we have analyzed the problem of context relevance and we have chosen an approach where the user is also involved also i...
Image Enhancement Using a New Shock Filter Formalism
We present a new approach based on Partial Differential Equations (PDEs) and shock filter theory for deblurring 2D Gaussian blurred images. The inherent problems of stability posed by the reverse heat equation and the di...