PROBLEM TO BE SOLVED: To solve the problem that the detecting accuracy of a biological signal is low in an utterance support device using a biological signal.SOLUTION: A person A who needs support is wearing a biological signal sensor part 1. The biological signal sensor part 1 is configured to detect the biological signal of the person A who needs support, and outputs a detected biological signal to a waveform shaping part 2. The waveform shaping part 2 is configured to output an analysis signal obtained by performing noise removal, interference removal, and waveform shaping processing to the detected biological signal to a word candidate detection part 3. An acquisition part 5 is configured of the biological signal sensor 1 and the waveform shaping part 2. A multi-word type word detection part is configured to detect a word sequence with high accuracy not only by using the feature information of each word but also by using feature information across a plurality of words with respect to the analysis signal, and to output the detected word sequence to an output part 4. The output part is configured to output the detected word sequence as an output device such as a speaker, a display part, a network connector (for example, computer), or a radio transmitter.COPYRIGHT: (C)2012,JPO&INPIT【課題】生体信号を用いた発話支援装置において、生体信号の検出精度が低いという課題がある。【解決手段】被支援者Aは、生体信号センサ部1を装着している。生体信号センサ部1は、被支援者Aの生体信号を検出し、検出後生体信号を、波形整形部2に出力する。波形整形部2は、検出後生体信号に対して、雑音除去、干渉除去、波形整形処理を行った解析信号を語候補検出部3に出力する。取得部5は、生体信号センサ1と波形整形部2で構成される。複数語型語検出部は、解析信号に対して、各語ごとの特徴情報を用いるだけでなく、複数語にまたがる特徴情報を用いる事によって、高精度に語列を検出し、検出後語列を出力部4に出力する。出力部は、スピーカー、もしくは表示部、もしくはネットワーク接続器(例えば、コンピューター)、もしくは無線送信器などの出力装置であり、検出後語列を出力する。【選択図】 図1
