Palmprint recognitionThe surface of the palm contains ridges and valleys just like fingerprints. In recent years, palmprint recognition has come to light because its template database is increasing. Palmprints pose certain advantages over fingerprints, such as the provision of additional features like wrinkles and principal lines, which can be easily extracted from images of slightly lower resolution. Palmprints provide more information than fingerprints, hence palmprints can be used to make an even more accurate biometric system. For palmprint recognition we will be exploring two methods: the first one [35] utilizes convolutional neural network-fast (CNN-F) architecture. It has eight layers: five convolutional and three fully connected ones. An experiment concerning the same over a PolyU palmprint database can be found in [35].
Another method for palmprint recognition comes from the research provided in [36]. This research utilizes a Siamese network model, and can be seen in Fig. 1.10. As explained earlier a Siamese network consists of two branch networks with shared weight. For the purpose of palm recognition, the two methods employ CNN networks, both with the VGG-16 model. Finally, instead of a simple fully connected network, we have a decision network that combines the layers in both branch networks. The architecture is composed of a dual five convolutional layer and three fully connected layers at the end, followed by a SoftMax.
Many important applications of pattern recognition can be characterized as either waveform classification or classification of geometric figures. For example, consider the problem of testing a machine for normal or abnormal operation by observing the output voltage of a microphone over a period of time. This problem reduces to discrimination of waveforms from good and bad machines. On the other hand, recognition of printed English characters corresponds to classification of geometric figures. In order to perform this type of classification, we must first measure the observable characteristics of the sample. The most primitive but assured way to extract all information contained in the sample is to measure the time-sampled values for a waveform, x(t1),…,x(tn), and the grey levels of pixels for a figure, x(1),…,x(n), as shown in Fig. 1-1. These n measurements form a vector X. Even under the normal machine condition, the observed waveforms are different each time the observation is made. Therefore, x(ti) is a random variable and will be expressed, using boldface, as x(ti). Likewise, X is called a random vector if its components are random variables and is expressed as X. Similar arguments hold for characters: the observation, x(i), varies from one A to another and therefore x(i) is a random variable, and X is a random vector.
