ECS Intranet:
Lightweight Signal Processing Algorithms and Architectures for Pervasive Healthcare

Recent advancements in wireless communication and wireless sensor network have opened up significant opportunity in futuristic healthcare and heath monitoring system development. Using sophisticated sensors and the wireless communication infrastructure monitoring a patient on a continuous basis will be feasible in the future. Typically such a system can be viewed as a layered structure consisting of 1) intelligent sensors layer, 2) network layer and 3) service layer. In the first layer a set of sensors (wearable or environmental) are deployed for collecting the vital signs of the patient under monitoring. These collected data are transmitted via the network layer to a central facility. WLAN/cellular/GSM/3G network could be used for serving this purpose. In the service layer the central facility processes these data to check for any abnormality of the received vital sign signals and accordingly the appropriate healthcare service is informed to take immediate action. Although very effective this structure faces the problem of transmission of huge amount of data over the network layer and effective management of the data at the central facility. The quality of transmitted medical data cannot be compromised by any means by using compression techniques since this may lead to false information about the condition of the patient under monitoring. Thus the main research effort has been dedicated to improve the quality and bandwidth in the network layer. Additionally maintenance of the data in the central facility needs intelligent database development. We envisage that a way out is to add intelligent processing circuitry in the sensor layer itself which can monitor the patient’s vital signs on a continuous basis and only transmit the signal to the central facility when it finds a departure of the vital sign’s pattern from the regular pattern. The raw medical data can be stored in a local computer using home wireless infrastructure. This approach reduces the burden on the network layer significantly. If required, the central facility can call for the raw data stored in the home computer over the network layer. However, processing the vital signs on a continuous basis is an extremely challenging problem in signal processing. Although there are several techniques developed for processing these data, they are very much computationally intensive and thus require significant amount of power. But in a wireless sensor network the biggest problem is the power and area. Thus these techniques are not likely to be suitable for embedding the associated circuitry within the sensor layer. To materialise our envisaged system, it is necessary to reduce the complexity of the required signal processing tasks (“light-weight signal processing”) and associated architectural optimisation in such a way that each of the processing circuit consumes as small power as possible. In the extreme case we envisage that the circuits may work through energy harvesting. Keeping this fact in mind we have initiated a research project for development of signal processing algorithms and associated ultra low-power architecture development for separating “a” vital sign signal from composite of similar kind of signals. This is a typical scenario when environmental sensors are used to monitor a patient who is visited by his/her relatives/friends. In this case the environmental sensor will receive a composite signal of several similar kinds of vital sign signals from which the circuit embedded in the sensor need to separate the only “necessary” signal corresponding to the person under monitoring. Since the person under monitoring can be mobile (within the room) it also needs to track the person on a continuous basis. The main challenge in this case is to find a clever algorithm of reduced complexity and associated ultra low-power architecture development. We would like to continue this work in this area since this leads to joint algorithm-circuit optimisation which eventually enables us to develop “real” pervasive healthcare system satisfying the stringent criteria of “pervasiveness”.

Funding

• EPSRC

Principal Investigators

• aa07r

Other Investigators

• Professor Koushik Maharatna
• Professor Bashir M Al-Hashimi
• Professor Steve R Gunn