In general, pattern recognition is the task of extracting
meaningful information from raw data that is typically
imperfect and noisy. After preprocessing, patterns can take
many different forms such as:
* The frequency distribution of sound (e.g. music, voice)
* The color histogram of images
* The gray level intensity of an image segment
* A string of ASCII code representing text
* A 2048 bit binary patternof a human iris
* Digitized signals from sensors
Any pattern is ultimately represented by a list of numbers -
parameters that describe the pattern. PME provides the
application critical task of recognizing or clustering
similar lists of numbers. It does not provide application
specific preprocessing and istherefore application independent.
General List of
Potential PME Applications:
The common factors are large databases or real time
processing - applications that require speed
From video cameras:
* Stereo Vision (multiple cameras for 3D vision)
* Optical flow (tracking movement of objects for vehicle
guidance)
* Robotic vision (guiding movement)
* Video Compression (for efficient transmission and
storage)
* Smart surveillance cameras (identify and track events
of interest)
* Object recognition (identifying objects of interest)
* Target tracking (smart munitions)
* Vehicle guidance (drones, robotics)
* Facial recognition, iris recognition (security -
biometrics)
From the Internet: *
Content-based image retrieval (searching for
similar images based on image content rather than by key
words)
* Search for music, video, voice by content
* Recognizing Malware signatures similar to known
threats
From Scanned Images:
* Forms processing (Forms recognition)
* Signature verification
* Character recognition (Postal code reading) * Satellite images (Recognizing objects of
interest – surveillance)
From existing databases:
* Data mining (Searching for meaningful patterns buried
in a massive database)
* Similar DNA sequence matching (from existing DNA
databases)
* Financial forecasting / stock picks / day trading
patterns (historical financial data)
* Detecting similar computer files (efficient archival
storage and virus detection)
From sensors
(biometric, audio, seismic, electromagnetic,
physiological, biomedical, tactile)
* Real time waveform recognition or compression
* Finger print recognition, voice identification
(security)
* Continuous speech recognition (automated
communications and transcription
* Word spotting (monitoring telephone communications -
security)
* Industrial process monitoring (detecting
anomalies/outliers – signal processing)
* Medical sensors (EKG monitoring, implantable sensors)
* From multiple sensors (audio, visual, tactile)