Big Data Multibiometric

Big Data Multibiometric: Pre-classification of Large Scale Multibiometric Database based on Light and Non-invasive Biometric Measures

One of the challenges in a biometric identification system is the performance when when dealing with large-scale databases. In an identification system it is necessary to search the entire database, being able to go through thousands of data records, and whose performance does not match with video monitoring systems of public spaces with great movement of people. One approach is to divide the recognition system into two subsystems, one that works with strong biometric characteristics and can use multiple biometric characteristics (fingerprint, face), and a second subsystem that acts on light biometries (sex, race, height , age, weight, hair color, skin color, eye color, walking style and tattoo). Strong biometrics are characteristics that determine the recognition of people in a unique way, for each individual, while light biometric measures deal with characteristics that provide some identifying information about an individual, but it is not enough to differentiate two individuals undoubtedly. Soft biometric systems are, in general, non-invasive, and can be extracted without the knowledge of the identified and at a distance. Before executing a recognition system based on strong biometric characteristics, it uses a process to identify light biometric characteristics and, thus, light biometric measures help to filter a large biometric database, and therefore, reducing the time of search. In addition to reducing the user’s waiting time, some studies show that multibiometric systems are also more robust. The proposal of this project is the development of a search filter system for the reduction of the biometric database based on light and non-invasive biometric measures. All lightweight biometric attributes are automatically extracted from security camera images.