Voice Activity Detection and Garbage Modelling for a Mobile Automatic Speech Recognition Application

摘要

Recently, state-of-the-art automatic speech recognition systems are used in various industries all over the world. Most of them are using a customized version of speech recognition system. The need for different versions arise due to different speech commands, lexicon, language and distinct work environment. It is essential for a speech recognizer to provide accurate and precise outputs in every working environment. However, the performance of a speech recognizer degrades quickly when noise intermingles with a work environment and also when out-of-vocabulary (OOV) words are spoken to the speech recognizer.This thesis consists of three different tasks which improve an automatic speech recognition application for mobile devices. The three tasks include building of a new acoustic model, improving the current voice activity detection and garbage modelling of OOV words.In this thesis, firstly, a Finnish acoustic model is trained for a company called Devoca Oy. The training data was recorded from different warehouse environments to improve the real-world speech recognition accuracy. Secondly, the Gammatone and Gabor features are extracted from the input speech frame to improve the voice activity detection (VAD). These features are applied to the VAD decision module of Pocketsphinx and a new neural-network classifier, to be classified as speech or non-speech. Lastly, a garbage model is developed for the OOV words. This model recognizes the words from outside the grammar and marks them as unknown on the application interface.This thesis evaluates the success of these three tasks with Finnish audio database and reports the overall improvement in the word error rate.