CAJ | 학술논문

EHMM人眼状态识别算法具有较高的识别率但算法复杂,因此利用SOPC开发平台设计了一种基于行列变换快速算法的2D-DCT IP核以提高处理速度;根据眼状态识别只需取2D-DCT后左上角部分子矩阵数据的特点,对其计算过程进一步优化;并引人了转置存储技术,浮点数乘法通过移位后转化为定点乘法器实现,优化了硬件资源,提高了处理速度;实验结果表明该IP核很好地实现了人眼图像灰度值的DCT变换.
EHMM인안상태식별산법구유교고적식별솔단산법복잡,인차이용SOPC개발평태설계료일충기우행렬변환쾌속산법적2D-DCT IP핵이제고처리속도;근거안상태식별지수취2D-DCT후좌상각부분자구진수거적특점,대기계산과정진일보우화;병인인료전치존저기술,부점수승법통과이위후전화위정점승법기실현,우화료경건자원,제고료처리속도;실험결과표명해IP핵흔호지실현료인안도상회도치적DCT변환.
Eyes'status identification algorithm based on EHMM(Embedded Hidden Markov Model)has high recognition rate but very complex,therefore,to enhance processing speed a 2D-DCT IP core with fast transform algorithm is implemented on SOPC platform.Since the eyes'status recognition only need the to Pleft corner partial matrix of DCT transformation,the computational process is further opti-mized.By the using of transposition storage technology and using the fixed-point multiplier instead of the floating number multiplication,the hardware is optimized and the processing speed is improved.The experiment results indicate it has achieved the anticipated goal very well.