局域网并行处理在语音识别中的应用

在语音识别中，不论是训练语音识别系统，还是使用系统识别语音，都需要进行大量的数据处理，这使得语音识别的研究和实现都变的非常困难。本文提出一种基于局域网的分布式计算机系统的快速并行数据处理方法来实现语音识别的模型训练和语音的识别，不仅加 了训练和识别的速度，节约了大量的时间，而且降低了语音识别任务对硬件的要求，取得了满意的效果。

THE APPLICATION OF LAN's PARALLEL PROCESS TO SPEECH RECOOGNITION

In the task of Speech Recognition (SR), no matter training a SR system or using a system to recognize speech, larger amount of data should be processed, which makes it difficult to the SR's research and realization. This paper presents a LAN-based fast data process method, distributed computer system parallel data process, to realize the SR's module train and the speech recognition, which not only can accelerate the speed of the train and the recognition, save a lot of time, but also bring down the request to the hardware. The experiment shows good result achieved.