The Hand Shape Recognition Method Research Based On Gaussian Mixture Model (GMM)

In the modern society, people's Identity Recognition is becoming more and more important. The traditional Identity Recognition methods are usually transformed into some objects which identify their personal identities, including keys, documents, ATM card, user name, password and so on. These traditional Identity Recognition methods have obvious disadvantages: individual possessions easily lost or forged, personal passwords easily forgotten or misremember. The more serious problem is that these systems are unable to distinguish genuine owners and the identity of the personaions that got the Identity Recognition objects. If others received these identity objects, they can have the same right. Moreover, the traditional Identity Recognition technology which has the characteristics of owning a password is becoming more and more difficult to meet the high security requirements, such as anti-terrorism, criminal investigation, information, and financial sectors. Biometric Identity Recognition technology is the technology that identify the physiological characteristics which the human body itself inherent through image processing and pattern recognition method automatically. For the biological characteristics themselves used for Identity Recognition carried by the users, it is not easy thieve and forgery. So it can compensate the shortcomings of traditional methods effectively. For the need of anti-terrorism, criminal investigation, information security, financial security and other areas, the research and application of biometric identification technology has drew more attention to all countries in the world unprecedently.The Hand Shape Recognition as one mainstream technology is becoming another important biometric identity method with its simple acquisition devices, fast certification after the fingerprinting biometric identification method. At present in foreign counties, in the commodity market share of biometric identification, the Hand Shape Recognition is basically in the same level with fingerprint identification. In 1996, in the Olympic Village which held the Atlanta Olympic Games, they used the Hand Shape Recognition system successfully. They accepted 65,000 people's registration and 1 million access control authentication. But literatures and samples at home and abroad on Hand Shape Recognition are very few. The Hand Authentication Technology is still in the research and development stage. In this paper, I mainly used Sony W1 digital camera to collect hand images and utilized the image processing, pattern recognition technologies and research methods comprehensively to achieve the image preprocessing, feature extraction, matching decision-making process and so on. I designed and completed a hand shape authentication system successfully and obtained good experimental results.In this paper, the important tasks and research results are as follows:(1)This paper introduced hand-shaped sample collection methods particularly. In the image segmentation part, according to the characteristics of threshold segmentation, I put forward a relatively new automatic threshold segmentation method. That is, in the gray image, different gray value has different pixel point. So the proportion is different. We prescribe a proportion T. All lower parts than the proportion are considered as low proportion area. Then we find out all the low proportion areas, and then identify the midpoints of this area and then find out the nearest mid-point that equal divides all the pixel point, which is the threshold segmentation point that we want to find out. Besides, we make use of the mathematical morphology method to extract the outline of hand-shaped curve. Then I use the method that curvature angle extracts angular point to isolate fingers, in which curvature angle is calculated by the geometric center supporting this region.(2)This paper adopted the Hand Shape Recognition Method based on Gaussian Mixture Model (GMM). In this paper, it only used three fingers'(index finger, middle finger, ring finger) information to verify your identity. And it used the two-thirds principle, that is, if two fingers pass the certification that the user can be certified. The basic idea of this method is to put the change of the finger shape as a random process that subordinate a certain probability distribution. We use the GMM to estimate the probability density function of the hand shape change.(3)In distinguish decision-making stage, we use Bhattacharyya distance as the distance metrics of the two corresponding fingers GMM.(4)I designed identity authentication system based on the hand shape recognition independently and fulfilled the function of preprocessing hand shape image, feature extraction and matching authentication. In the image acquisition stage, it adopted digital camera as a collection end, not too many restrictions for the hand Movement. The operation is relatively simple, more humanization, easier to acceptance. Aiming at our established small hand shape base, we did an effective verification on the preprocessing of hand shape image, feature extraction and recognition algorithm and got specific test results. The experiment results show that the algorithm presented in this paper has a good robustness, feasibility and accuracy.