Distributed Speech Recognition System for PDA in Wireless Network Environment

In this paper, we proposed a distributed speech recognition system using PDA (Personal Digital Assistant) as a client and personal computer as a server through wireless network. General mobile clients such as PDA’s are primarily motivated by the needs for providing more convenient user-interface. Speech or character recognition is provided for this purpose. However, the conventional mobile client is hard to employ such user friendly interfaces because of its limited processing power and its limited memory space. To solve the problem it is considered that a client transmits the extracted features from the inputs (voice, character, etc.) to the server, the server return the recognized results to the client, and the client display them to the user. By doing this way, a larger size of recognition engine can be used for better recognition accuracy. HMnet (Hidden Markov network) models are widely used as the basic units for recognition of continuous speech, isolated words, on-line characters, and so on. We will discuss the architecture of proposed distributed system and several constraints that must be considered in the server of the system. Then we discuss multiple connection processing, sequential processing using data queue, and adapted model in each client’s environments to solve these constraints.