Earticle

초록

영어

The feature extraction process is the fundamental stage in the development of speech processing systems. In this paper is described the methodology and implementation of an embedded algorithm for extracting characteristic information from silent speech signals. The classical approach based on frequency representations of the signals is followed as methodology for this work. First, the acquisition stage of the silent speech signals is performed using a Non-Audible Murmur microphone and the STM32F4Discovery evaluation board. Then, the digitalized signal is filtered, segmented and normalized using the pre-emphasis and windowing steps. The magnitude spectrogram is calculated from the pre-processed signal using the Fast Fourier Transform (FFT), to finally estimate characteristic data from de silent speech signal. As result of this process, the signal characteristic parameters, defined in the frequency domain are obtained and used as elements at later stages of pattern recognition, in order to build systems of Automatic Speech Recognition (ASR), Speech Coding, Speaker Recognition, among other applications.

목차

Abstract
 1. Introduction
 2. Methodology
  2.1. Acquisition of NAM Signals
  2.2. Pre-emphasis
  2.3. Signal Windowing
  2.4. FFT Calculation
 3. Results
 4. Conclusions
 References

키워드

저자

• Leonardo A. Góngora [ Mechatronics Engineering Program, Faculty of Engineering, Nueva Granada Military University, Bogotá, Colombia ]
• Dario Amaya [ Mechatronics Engineering Program, Faculty of Engineering, Nueva Granada Military University, Bogotá, Colombia ]
• Olga L. Ramos [ Mechatronics Engineering Program, Faculty of Engineering, Nueva Granada Military University, Bogotá, Colombia ]

참고문헌

발행기관

간행물

표지보기 관심저널 등록

• 간행물명

  International Journal of Multimedia and Ubiquitous Engineering

• 간기

  월간

• pISSN

  1975-0080

• 수록기간

  2008~2016

• 등재여부

  SCOPUS

• 십진분류

  KDC 505 DDC 605