Nhận dạng tiếng Việt nói sử dụng bộ công cụ Kaldi

Abstract

Speech recognition has been increasingly applied in various fields such as automatic switchboards, security, searching  by voice… however the quality of recognition is the problem of utmost concern. The Kaldi toolkit is a new tool developed in 2009. Kaldi was introduced at a workshop  held at Johns Hopkins University with the title “Low Development Cost, High Quality Speech Recognition for New Languages and Domains”. This paper describes the Vietnamese speech recognition system built on Kaldi toolkit. The paper also evaluates quality of the system based on the evaluation the ratio of the WER on AMs (Acoustic models). The system has superior results compared the previous toolkit to Vietnamese speech.