- 追加された行はこの色です。
- 削除された行はこの色です。
[[HOME/RESEARCH TOPICS]]
* 研究内容 - 音声認識 [#xf67d5d7]
* RESEARCH TOPICS - SPEECH RECOGNITION [#k7937f36]
** はじめに [#d2dc5acb]
** Introduction [#d2dc5acb]
> 音声認識とは，人が話した音声をコンピュータに理解させて文字データにすることです．~
この技術によって，音声でロボットやコンピュータの操作を行うことができます．
> Speech recognition is a processing which extracts character data by recognizing utterances.~
By this technology, we can operate robots or computers by our voices.
** HMM音声認識 [#a0ef36c7]
** HMM speech recognition [#a0ef36c7]
>Generally, speech recognition is studied separately on acoustic models with acoustic characteristics and language models with language characteristics.~
HMM Speech recognition is a technology of speech recognition which studied in our laboratory well, and it models acoustic models with based on probability models called HMM(Hidden Markov Model).
> 音声認識は一般的に，音響的特徴をもつ音響モデルと言語的特徴をもつ言語モデルの2つについて別々に研究されます．~
本研究室でよく研究されている音声認識技術であるHMM音声認識は，音響モデルをHMM(隠れマルコフモデル)と呼ばれる確率モデルに基づいてモデリングしています．
** Research Field [#a783d056]
** 研究分野 [#a783d056]
*** Sequential Speech Recognition [#j6f1afbe]
*** 連続音声認識 [#j6f1afbe]
> This technology analyzes speeches and converts to sentences.~
This is a research for getting better speech recognition accuracy.
> 音声を解析して文章に変換します．~
より高い音声認識精度を得るための研究です．
*** Speaker Adaptation [#j6f1afbe]
>The property of human voices have a great deal of characteristics in sex or age, so it is difficult for us to recognize it at high accuracy, however, we hope more capable system by specializing in speech recognition system applied to particular target.~
This is a research in terms of specializing in systems for particular speaker.
*** Speaker Identification [#j6f1afbe]
>This is a technology that identifies whose voice input sounds are or judges whether preregistered speakers or not.~
This is a research for using automatic speech archives for each speakers at security systems or meetings, and so on.
