精简AlexNet的指静脉识别

Finger Vein Identification Using Simplified Alex Net

指静脉识别技术是目前公认的最安全的生物识别技术,拥有活体识别、内部特征、非接触式采集等优势。指静脉识别技术在银行金融行业、保险箱、社会保障等领域等对安全性要求高的领域有着巨大的需求。本文针对AlexNet在指静脉识别中训练时间长、卷积视野较大等问题,提出了精简的AlexNet模型。为了加快网络的收敛速度,使用Xavier对精简后的AlexNet进行初始化。实验结果表明,精简后的模型在公开和自有指静脉数据集上的准确率上有所提升。

Finger vein identification is known as the safest biometric identification technology with advantages of living identification,internal feature,non-direct capture,etc.Finger vein identification is in great demand in certain areas that require high safety,such as banking,finance,safe deposit box,social security,and so on.A simplified AlexNet model is proposed to solve the problems such as long training time and large convolutional visual field in finger vein identification.In order to accelerate the convergence speed of the network,Xavier initialization was introduced to initialize the simplified AlexNet.Experimental results show that the simplified model is more accurate on the open and private finger vein data sets.