digital biomedical signal processing
adaptive computing systems
In the paper, high-reliability computers based on adaptive computing systems are considered. Such computers can be used in electronic medical equipment to process different biomedical signals including one-dimensional and multi-dimensional signals (e.g., electrocardiogram, images, and video data). As the first step of this research, the methods to increase the reliability of computers on their design stage are classified and analyzed. The efficiency of the traditional methods oriented on the parallel usage the same electronic component is estimated. It can be insufficient for important applications like medical equipment.
The idea behind the proposed approach is to use different but functionality-similar hardware. For instance, digital signal processors (DSPs), field-programmable gate arrays (FPGAs) and general-purpose processors (GPPs), which contain additional digital signal processing instructions such as extended multiply ones, can be used for this aim. The paper presents mathematical proofs why the reliability of such a triple system is higher than the reliability of electronic systems exploiting traditional reservation methods (e.g., parallel usage of the same electronic compo-nents). In addition, a structure scheme of a high-reliability dynamically reconfigurable computer and its state ma-chine is considered in detail. The mathematical tools to estimate the reliability of this adaptive computing system are presented.
Another key issue is the reliability of software used in such computers. Since software errors are an objective problem, which cannot be avoided fully, it is offered to use different real-time operating systems in order to exclude essential software problems such as starvations. For the sake of the paper completeness, different operating systems including cross-platform (e.g., QNX, LynxOS, VxWorks) and application specific (e.g., DSP/BIOS for DSPs developed and manufactured by Texas Instruments Inc.) ones are shortly overviewed.