Physiological signals are those from sensors placed on the body or implanted. Although a challenge to interpret, several physiological signal-processing (PSP) systems have been developed. Some, such as electrocardiograms and electronic stethoscopes, only gather information relevant to the assessment of a physiological system. Another classification preprocesses sensory signals to compensate for a deficiency in peripheral systems. Examples include hearing aids, cochlear implants, and vision-assist devices.

And a third type of PSP system interacts closely with the physiological system, becoming a part of the physiological control loop. This is a relatively new and emerging class including cardiac-rhythm management (CRM) systems, automatic drug-delivery (ADD) systems, and smart artificial limbs interacting with neural signals. For example, CRM systems help control cardiac activity by monitoring and analyzing electrical and mechanical activity, and providing therapy in the form of electrical stimulus when needed.

In this latter class, physiological information is converted from analog to digital signals. Raw data is often transformed to an alternate domain (such as a frequency or time-frequency domain) where important info can be extracted.

Such systems are either implanted in the human body or carried by the user, so power consumption and sometimes physical size are seriously limited. These requirements limit the scope of signal processing in PSP systems. Yet, the opportunity for significant benefit exists precisely because PSP systems operate in real-time synchronization with the underlying physiological system.

One way to combat these restrictions is to use real-time PSP systems operating in the time-frequency domain provided by Weighted Overlap-Add (WOLA) filterbanks targeting low-delay and implantable devices. WOLA-based processing has several benefits:

• Efficient decomposition of signals into orthogonal signals, each in a different frequency band, or subband. As a result, there's more flexibility to use different strategies and parameters in individual or groups of subbands.

• Lower delay and processing latency compared to standard frequency-domain methods.

• Superior representation. Because of orthogonal subband decomposition, various physiological events are easier to detect in individual or groups of subbands.

• Parallelism in signal processing. This will eventually lead to stronger and faster processing and decision-making.

• Although the advantages of subband processing have been known for years, algorithmic complexity that leads to high computation cost and formidable power consumption prevented their use in most implantable devices. The WOLA-filterbank hardware platform provides the opportunity to use sophisticated subband-based signal processing techniques. The filterbank also offers a flexible configuration easily optimized for processing delay, frequency resolution, and computation cost for the desired application.

Apart from numerous successful audio and speech applications, subband-based signal processing has been used for many PSP applications as well. Some examples are heart sound analysis in electronic stethoscopes, EEG analysis for event detection, epilepsy seizure detection and prediction.