The established single-biometric recognition systems which are fingerprint, face, speech, iris, and what not have many problems of FAR(False-Acceptance Ratio) and FRR(False-Rejection Ratio). This research studied embedded multi-modal biometrics recognition system using unique features of voice and fingerprint in order to increase reliance and complete that problems.
The whole system that was implemented divided into three parts, fingerprint recognition part, speaker recognition part, control part. First, voice that is inputed from microphone extract features of voice through pre-emphasis, hamming window, MFCC(Mel-Frequency Cepstrum Coefficient) after voice area was detected by short time energy. The extractive features of voice area are stored to database and achieves recognition process by DTW(Dynamic Time Warping) algorithm using voice features at speaker recognition process. Second, fingerprint image that is inputed from AFS-8500 that is a semi-conductor fingerprint sensor achieves LPF(Low Pass Filter), histogram equalization, binarization, thinning process and makes direction pattern and detects singular point of fingerprint. After this process is completed, it achieves feature points detection of fingerprint and post-process. It is trained feature points and singular points of abstracted fingerprint using autonomic neural network KSOM(Kohonen Self Organizing Maps) that is one of Artificial Intelligence techniques. If training process of all feature data is completed, KSOM achieves recognition process about fingerprint. Third, the control part controls whole action of fingerprint recognition part and speaker recognition part.
As a result of this research, the performance of this research dropped quantity in whole recognition rate than single recognition system. However, the performance is displaying that improved much about FAR that is the most important urea of recognition system. Finally we may have to develop recognition system that uses various biometrics information and increase stability more for research direction hereafter.