About Me
I received the Ph.D. in Computer Science from Michigan State University in May, 2012.
I am now a Lead Scientist at
Noblis
where I am continuing my research in face
recognition, biometrics, and pattern recognition.
What is Pattern Recognition?
Pattern recognition is the categorization or separation of signals and data. As humans we do this on a constant basis: from recognizing a friend to distinguishing between a road and a field. While we do not give much thought to these tasks, programming a computer to perform the same feats is far from trivial. Pattern recognition research is most often conducted in computer science departments these days, though it is the marriage of signal processing (electrical engineering), statistics, linear algebra (mathematics), algorithmic computer science, and other fields.
The signal processing aspect of pattern recognition typically involves the initial processing of a signal (pattern sensing). One common signal used in pattern recognition is digital images. Extracting information from these images is known as computer vision, which has evolved from signal processing. Signal processing and computer vision algorithms seek to compensate for noise (inaccurate pixel intensities) in images, reduce the dimensionality of a signal by projecting it into a alternate basis (Fourier transform, cosine transform, principal component analysis, feature extraction etc.), and extract image features for organization and classification.
The statistical aspect of pattern recognition generally consists of a learning phase and a decision phase. These algorithms are also part of the field of machine learning. The learning phase (generally) uses statistical methods learn a model of the problem being addressed. The decision phase then applies this model to a newly acquired signal. Classically machine learning is dichotomized into supervised and unsupervised learning. Supervised learning consists of patterns (extracted during the signal processing phase) that have a known category or label. The key to the learning phase is that data (or observations) are needed, which can be highly expensive to collect.
What is Biometrics?
Biometrics is the method of using characteristics of the human body for the purpose of determining or verifying a person's identity. Typically this involves measurements or assessments of specific parts of the body (single biometrics), though this can also involve the use of multiple regions of the body (multi-biometrics). The three most popular biometrics are fingerprint, face, and iris. Other biometrics that have been used successfully include, and are not limited to, hand, palm, ear, vascular, speech, gait, and dental. Biometrics are an applied field of pattern recognition, where the sensed data is generally digital images (e.g. scanned fingerprints, face photographs, infrared iris images) and the algorithms used generally measure the distance of subjects in a some geometrical space (created from feature-based representations, linear subspaces, etc.).