Portable computers for real-time signal processing: EEG analysis as a case study

Recent advances in both digital hardware and digital signal theory have led to a rapid expansion in the importance and application of computer-aided measurement (CAM) techniques. Of these advances, the emergence of cheap microprocessor technology of sufficient processing power and speed to support some of the real-time signal-processing tasks encountered in CAM, is probably the single most important factor. The Roving Slave Processor (RSP) represents a novel extension to the field of CAM. The RSP is a basic hardware unit comprising, in its simplest form, a central processor, a memory system and a means' of input-output. By the use of a microprocessor, it is possible to reduce the size of the complete system to very small dimensions, i.e. to construct a portable computer. The unit is wholly dependent upon a master computer for the provision of all fundamental peripherals (e.g. teletype, reader-punch, etc.) and for all program preparation. To provide these facilities, a special purpose interface has been constructed. The RSP is, however, capable of disconnected operation and this is shown to lead to a very efficient and economical means by which to perform CAM operations. The design and development of two prototypes is described with particular attention being given to the choice of processor, the storage system and the link to the rnaster computer. Some consideration has also been given to the problem of how the RSPs should be programmed and a scheme based on a high-level calling system is detailed. Problems of reliability, both hardware and software, are also discussed. An application of the RSP technique in the very demanding field of real-time EEG analysis is described, with particular attention given to the development of an automatic spike detector algorithm. The occurrence of spikes in the EEG signal is of particular clinical significance as it is indicative of the onset of an epileptic attack. Sharpwaves, slow-waves and all other abnormal behaviour have been omitted from this study. A system, based on a filtered, first-order difference of the EEG signal has been developed and is described. Very encouraging results have been obtained, with a 95% success rate for the abnormal spikes occurring in a series of test records. Finally, techniques for the production of a miniature version of the RSP, which may be attached to and conveniently carried by a patient, are discussed.