結合音長與發音特徵於GTB之腔調化語音辨識

Accented Speech Recognition based on Gradient Tree Boosting with Duration and Articulation Features

現在利用語音作為輸入的人機介面上，須要考慮到發音變異對語音辨識（speech recognition）所造成的影響。通常是以語者或語音調適（speaker adaptation and speech adaptation）的技術來處理。但是對有腔調（accent）之語者的語音辨識效果較不理想。本論文提出將兩個聲學模型（phone acoustic model）進行合併，來重建聲學模型，來降低腔調化語音的錯誤率。首先建立帶有腔調化特徵之聲學模型，合併到傳統聲學模型中，合併時，使用狀態聯繫（state tying）與梯度漸近樹（gradient tree boosting, GTB）演算法，來調整傳統聲學模型，以重建聲學模型。以便提升腔調話語音的辨識能力。

Using speech as the input of the human-machine interface, we need to consider the effect of the pronunciation variations on speech recognition. We usually use the speaker adaptation and speech adaptation technique to solve above problem, but the result of accent speech recognition is not good enough. This paper presents a framework to combine two phone models to reconstruct acoustic model. First we build accented acoustic model and unaccented acoustic model, than we combine the accented acoustic model into the unaccented acoustic model with the state tying and Gradient tree boosting algorithm. We can adjust the unaccented model by this framework, to reconstruct acoustic model and robust the accented speech recognition performance.