A HMM Speech Recognition System Based on FPGA

This paper introduces an embedded speech recognition system, including both hardware and software frameworks, based on an FPGA platform and the HMM algorithm. The paper also proposes an approach to the design and implementation of an IP core for the forward algorithm, as well as simulating the algorithm. The main advantage of this approach is the reduction of processing time in speech recognition because the algorithm is implemented by dedicated hardware. The experimental result shows the system performs well and is helpful for speech recognition research based on an IP core in an embedded platform.