Segmentation method strings of words in the training phase of a recognizer connected words.

摘要

CADENAS segmentation method WORDS IN TRAINING PHASE recognizer connected words. TYPE OF comprising the steps segmenting Automatics (2) CHAIN ​​OF WORDS IN PROGRESS A SEGMENT (1) COMPARISON (3) OF THE NUMBER AND IDENTIFICATION OF WORDS THAT HAS BEEN TARGETED THE CHAIN ​​OF WORDS IN COURSE SEGMENT (1) WITH THE VALUES OF THE NUMBER OF WORDS AND IDENTIFICATION OF RELEVANT TO THE ORIGINAL CHAIN ​​OF WORDS (4) thereof, and Decision (5) ON THE INCORPORATION (8) segmentation OBTAINED THE HIDDEN MARKOV MODELS WHEN COMPARISON (3) SHOWS THAT THE NUMBER OF WORDS AND THE IDENTIFICATION OF THEM ARE CREATED EQUAL. The method is characterized in that when the above comparison shows that the values ​​MENTIONED ARE NOT CREATED EQUAL, IS A PROPORTIONAL SEGMENTATION (6) CHAIN ​​OF WORDS IN PROGRESS A SEGMENT (1), comprising dividing said CHAIN ​​IN MANY SEGMENTS AS WORDS CONTAINING THE WORD CADENAORIGINAL (4) and assign each SEGMENT DURATION TO THE DURATION OF PROPORTIONAL corresponding word when pronounced alone (7).