Loss-Resilient Sensor Networking monitoring in mobile health apps - Embedded.com

Loss-Resilient Sensor Networking monitoring in mobile health apps

Many elder patients have multiple health conditions such as heart attacks (of various kinds), brain problems (such as seizure, mental disorder, etc.), high blood pressure, etc. Monitoring those conditions needs different types of sensors for analog signal data acquisition, such as electrocardiogram (ECG) for heart beats, electroencephalogram (EEG) for brain signals, and electromyogram (EMG) for muscles motions.

To reduce mobile-health (m-health) cost, the above sensors should be made in tiny size, low memory, and long-term battery operations. We have designed a series of medical sensors with wireless networking capabilities.In this paper, we report our work in three aspects:

(1) networked embedded system design,(2) network congestion reduction, and (3) network loss compensation. First, for networked embedded system design, we have designed an integrated wireless sensor network hardware / software platform for multi-condition patient monitoring.

This work proposes a novel approach of interfacing medical sensors to a RF mote (which includes wireless communication and a Microcontroller) through a highly versatile 8051-based PSoC1 Mixed-Signal array from Cypress which integrates ECG/EEG/other sensors with Radio Frequency Identification (RFID) into a Radio Frequency (RF) board through a programmable interface.

For network congestion reduction, the interface chip can use compressive signal processing to extract bio-signal feature parameters and only transmit those parameters. This provides an alternative approach to sensor network congestion reduction that aims to alleviate “hot spot” issues.

For network loss compensation, we have designed wireless loss recovery schemes for different situations as follows:

(1) If original sensor data streams are transmitted, network congestion will be a big concern due to the heavy traffic. A receiver-only loss prediction will be a good solution.

(2) If the signal parameters are transmitted, the transmission loss mandates a 100% recovery rate. We have comprehensively compared the performance of those schemes. The proposed mechanisms for m-health system have potentially significant impacts on today’s elder nursing home management and other mobile patient monitoring applications.

(3) For compressed transmission, each coefficient is important from signal reconstruction viewpoint. TCP cannot be used although they can achieve 100% recovery. Therefore, we proposed a ripple-based local recovery and erasure-codes-based approach to guarantee all data’s safe arrival.

(4) The sensor network based telehealthcare system could be integrated with RFID technology to achieve patient medicine-taking monitoring. We have built reprogrammable RFID reader for future security and communication protocol enhancement purposes.

The use of the PSoc extends the computational capabilities of the RF mote, while keeping its power consumption in check. The PSoC's mixed signal array allows programming of analog and digital (mixed-signal) components that are typically used in embedded systems.

It also has a built-in microcontroller which integrates and controls all of the programmed components. Because of the extended computational capabilities through using this mixed signal array, complex computations such as filtering or triggering as well as application-specific data compression or suppression can be implemented at individual nodes. This can further reduce the data throughput over the network and result in reduced transmission time and improved network traffic.

Our PSoCc interface has used wavelet-based signal decomposition to obtain the signal feature parameters (coefficients). Those features can also be used for normal/abnormal signal classi- fication and normal signal will not be transmitted.

From wireless loss recovery viewpoint, if no 100% loss recovery is required (such as in uncompressed ECG stream transmission case), a destination-only particle filter or EKF’s loss prediction / compensation could work efficiently from our above discussions.

To read this external content in full, download the complete paper from the open online author archives at the University of Alabama.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.