Analytic model of sentence streaming using TCP for speech recognition

An implementation scheme of speech recognition over Internet called network speech recognition (NSR) can use TCP to ensure the reception of speech data on the server so that the accuracy of recognition obtained could be better. Although TCP can not guarantee the timeliness of data delivery as required by multimedia streaming applications, NSR prefer to receive the whole speech data to be recognized on the server rather than delay. In this paper, we enhance our study [1] about an analytic model of NSR recognizing a speech sent via TCP sentence by sentence by realizing and solving the model with the help of modeling tool TANGRAM-II.