Title :
A Front-End Solution to VAD, SNR, and AGC for Speech Recognition
Author_Institution :
Depart of Math., Zhejiang Univ. of Technol., Hangzhou, China
Abstract :
In this paper VAD, SNR and AGC, as main complications of a front-end signal processor in automatic speech recognition are discussed with their structure and realization, including experimental evaluation on the module global-SNR is given. The front-end has five advantages: i) propose a novel scheme to long-term power estimation based on accumulate average; ii) elegant policy for VAD based on pitch detection; iii) relative highly accurate SNR estimation; iv) AGC based on soundcard when silence between utterances; v) robust performance even in very noisy acoustical circumstance.
Keywords :
audio signal processing; speech recognition; automatic speech recognition; front-end signal processor; front-end solution; noisy acoustical circumstance; Acoustic noise; Acoustic signal detection; Automatic speech recognition; Information science; Mathematics; Robustness; Signal detection; Signal processing; Signal to noise ratio; Speech recognition;
Conference_Titel :
Information Science and Engineering (ICISE), 2009 1st International Conference on
Conference_Location :
Nanjing
Print_ISBN :
978-1-4244-4909-5
DOI :
10.1109/ICISE.2009.37