A New Word Spotting Framework Using Hough Transform of Distance Matrix Images

摘要

In this report, a new word spotting framework using Hough transform of distance matrix images is proposed The brightness of an intersectional pixel in the distance matrix image expresses the distance value between an input and a standard pattern If a standard pattern word is included in the input utterance, the straight line area is observed in the distance matrix image If the voting number for the corresponding Hough transform parameters is more than the threshold value, it is decided that the standard pattern word exists in the input utterance. This new framework was evaluated by the recognition accuracy, the correct/incorrect words rejection rate and the false alarm.