Speech recognition device and speech recognition method

Abstract

The speech recognition apparatus ( 1 ) is equipped with the garbage acoustic model storage unit ( 110 ) storing the garbage acoustic model which learned the collection of the unnecessary words; the feature value calculation unit ( 101 ) which calculates the feature parameter necessary for recognition by acoustically analyzing the unidentified input speech including the non-language speech per frame which is a unit for speech analysis; the garbage acoustic score calculation unit ( 111 ) which calculates the garbage acoustic score by comparing the feature parameter and the garbage acoustic model; the garbage acoustic score correction unit ( 113 ) which corrects the garbage acoustic score calculated by the garbage acoustic score calculation unit ( 111 ) so as to raise it in the frame where the non-language speech is inputted; and the recognition result output unit ( 105 ) which outputs, as the recognition result of the unidentified input speech, the word string with the highest cumulative score of the language score, the word acoustic score, and the garbage acoustic score which is corrected by the garbage acoustic score correcting means.

Description

technical field [0001] The present invention relates to a speech recognition device and a speech recognition method for speech recognition of continuous word speech recognition of useless words that do not need to be distinguished in the allowable sense. Background technique [0002] In the past, there is a word speech recognition device that uses a sound model learned in advance from a collection of useless words—a useless information sound model to deal with useless words that do not need to be distinguished in meaning (for example, please refer to 2 people such as (Japan) Inokami Naoki, " ガ一ベジHMMを用いた自由虺虺のの广语设计技术(Using Useless Information HMM in Natural Speech Sentences to Process Useless Words)”, Journal of Electronics, Information and Communications Society A, Vol.J77-A, No.2 , pp. 215-222, February 1994). [0003] figure 1 It is a configuration diagram showing a conventional speech recognition device. [0004] Such as figure 1 As shown, the speech recognition device ...

AI Technical Summary

Problems solved by technology

[0025] However, in the above conventional example, there is the following problem: if there is a word sequence similar in sound to non-linguistic sounds such as (stammer) sounds in the vocabulary to be recognized, it will be misrecognized.

[0029] The reason for this is that the useless information sound model is learned from all the sound data of useless words including eating sounds, so the distribution is very wide, and a high sound score cannot be obtained for the pronunciation of useless words, that is, non-linguistic sounds

[0030] As a solution to this, there is a method of uniformly increasing the garbage sound score, but this method also increases the value of the garbage sound score in frames where the optimal path is not a garbage word, so it becomes a cause of misrecognition

Images