Establishment of Japanese Continuous Speech Recognition System Based on Association Rules Mining

Abstract

Voice technology is to make these devices and machines recognize and understand correctly through a series of processing of human voice signal information. With the increasing application demand of speech recognition, speech recognition technology is gradually becoming the key technology in human-computer interaction. In this paper, the technology based on ARM (Association Rule Mining) is applied to the establishment of Japanese continuous speech recognition system. Considering the characteristics of Japanese hiragana, katakana and Japanese Chinese characters, the effects of different modeling units on recognition performance are discussed through experiments on Japanese data sets. On this basis, Apriori algorithm and FP-growth algorithm are used to extract speech emotion prosodic features, and the advantages and disadvantages of these two algorithms are analyzed and compared from the number of effective rules generated by the algorithm and the running time, and finally Apriori algorithm is selected to extract speech emotion prosodic features.