Convolutional neural network speech recognition system based on FPGA

The invention relates to a convolutional neural network speech recognition system based on an FPGA (Field Programmable Gate Array), which comprises software implementation and effect of a convolutional neural network (CNN) on speech recognition and research on a CNN speech recognition technology based on the FPGA, and realizes a set of complete speech recognition system by fully utilizing the advantages of software and hardware co-processing. The effect of a CNN network in a speech recognition acoustic model is researched, modeling research is carried out for different people, different regions and various dialects, how to realize implementation and acceleration of a CNN convolution model algorithm by using an FPGA is researched, and a series of processes of preprocessing, feature extraction, model prediction and the like of speech recognition on the FPGA are provided. The characteristics of high parallelism, high throughput rate and low power consumption of the FPGA are fully utilized, and the speech recogni