Most of current palm vein recognition systems use an active light source to acquire images. The choice of light-source wavelength directly affects the definition of palm vein image and recognition performance. Typical wavelength of palm vein recognition is 760, 850, 890, 940 nm, while few works have been done on investigating which one is the optimal wavelength for palm vein recognitions. It is solved from two angles. From the angle of feature extraction, the model of palm vein image definition is established and palm vein definition of the four wavelengths is compared according to the model. From the angle of feature matching, the recognition performance of palm vein image with four wavelengths separately by three typical biological identification algorithms is compared. The experiment was done in a self-build database including 2400 palm vein images with four kinds of wavelength. The experiment result of the model and 3 algorithms show that 850 nm is the optimal wavelength. Wavelength of 850 nm achieves higher palm vein recognition performance than the other three wavelengths.