人工神经网络具有强大的模式分类能力,除此之外还有一个优点即其大量的并行神经计算能力,使得以低廉的硬件开销来实现高速并行运算成为可能,这在实时模式识别中是非常重要的。文章介绍了以预言神一号小型神经计算机的并行神经运算为核心的连续语音的实时识别的算法,实验表明该算法有很高的实时性和较好的识别率。 Artificial neural network has a strong model classification ability, in addition to an advantage that its large amount of parallel neural computing power, making low-cost hardware to achieve high-speed parallel computing possible, which is very important in real-time pattern recognition of. In this paper, an algorithm for real-time speech recognition based on predictive parallel neuronal computing of God 1’s small neural computer is introduced. Experiments show that this algorithm has high real-time performance and good recognition rate.