WORD SPOTTING VOICE RECOGNIZING METHOD

PURPOSE:To improve the recognition rate, specially, accumulated recognition rate by outputting a compound word as a 1st candidate. CONSTITUTION:A voice recognition part 1 recognizes partial words, compound word (word containing partial word), and difference words (word corresponding to difference between partial word and compound word) respectively and rearranged the recognized words in recognized likelihood decreasing order, and the recognition results are stored temporarily in a recognition result storage part 2. When a partial word comes first as a recognition result, a candidate reselection part 3 extracts a compound word whose recognized likelihood with the 1st partial word is smaller than a specific value among compound words containing the partial word, finds the difference word of the extracted compound word which has recognized likelihood larger than the specific value, and outputs the compound word containing the difference word as the 1st recognition result word when the word is detected.