Electroglottograph as an additional source of information in isolated word recognition

Abstract

Traditionally, speech recognition systems use only the acoustic speech signal (speech). However, the source of the signal and the way speech is produced and whether this information can aid in speech recognition needs to be investigated. The objective of this study was to assess the contribution of using the electroglottograph (EGG) as an additional source of information along with speech in an isolated word recognition system. The vocabulary consisted of 64 words, ranging from mono-syllabic words to words with four syllables. Two fully connected artificial neural networks were designed. One network (speech network) used only speech as its source of information. The other network (speech+EGG network) used EGG along with the acoustic speech signal as its source of information. The speech network had a peak recognition rate of 94.37%. The speech+EGG network had a peak recognition rate of 99.37%. Hence, the information provided by the EGG improved the performance of the speech recognition system by 5%.