A fast palmprint feature extraction method based on embedded device

作者：Wang Jiaqiang; Yu Ming; Qu Hanbing^*

来源：Revista de la Facultad de Ingenieria, 2017, 32(1): 752-756.

• 2-D Gabor filter
• Competitive coding
• Embedded device
• Gabor function
• Palmprint feature extraction
• Palmprint identification
• Real time requirement
• Recursive implementation

This paper proposes an optimized competitive coding method for palmprint identification. A recursive implementation Gabor filter is used instead of the Gabor function based neurophysiology. The recursive implementation Gabor filter decompose 2D Gabor filter into two nonorthogonal 1D Gabor filters, x-direction and φ-direction,and then calculate the point and the three points before and after in each direction. Experimental results show that this method, compared with the competitive coding method,EER increased by only 0.0004%. But the speed increased significantly, image with size of 128 128 feature extraction time reduced from 330ms to 160ms in the ARM9 platform and reduced from 63ms to 30ms in the Cortex-A8 platform, so this method is fully applicable to embedded devices, especially for high real-time requirements.

• 出版日期2017
• 单位河北工业大学