The IP core design of eye states recognition based on EHMM

The eye states recognition based on EHMM (Embedded Hidden Markov Model) has a high recognition rate but is complicated. By using the on-chip memory and the parallel characteristic of the hardware logic, the algorithm was implemented into IP core in this paper. Through the deep analysis of the data stream, a parallel pipeline was proposed. Then IP cores of sub-modules were designed. At last, the eye states recognition IP core was integrated on the SOPC. The results show that it takes only 60ms to detect an eye image with a 50Mhz CPU. The speed reached the real-time requirements and was 170 times faster than the software implementation. The structure proposed in this paper working on the SOPC system at low frequencies performs as well as the PC working at high frequencies.