Techniques to Implement an Embedded Laser Sensor for Pattern Recognition

This paper describes the implementation of a real-time, non contact, static hand sign recognition system using simple techniques and cost effective equipment. The system has two operation modes: recognition mode and learning mode. Hand sign recognition is based on an image appearance model of the human hand extracted by means of a custom transducer. The transducer is composed of a Web cam and a laser plane generator. A DSP processes two images, extracts a laser trace and computes the AR coefficients which are fed to an MLP neural-network classifier. The result of the classification operation is translated into a command sent to the target system (e.g. a PC). The default set of recognized signs can be customized in learning mode by "tuning" the system according to needs. The rates of correct recognition for all testing hand signs were in the range of 0.82 -1