Embedded signal processing for WBSN applications

***Archived: This is a past project,  it is not on offer at the moment!***


Project description:

In the last few years, Wireless Body Sensor Networks (WBSN) have emerged as the solution for the future generation of healthcare, providing ambulatory, continuous and real-time monitoring of vital signs. Typically, a WBSN is composed of a bunch of sensor nodes that sense various vital signs such as ECG, EMG, body temperature, blood pressure and so on. Each sensor node is composed of a limited number of components including: an analog readout front-end, a microprocessor, a radio transmitter/receiver and a power alimentation circuit together with batteries. These sensor nodes then transmit their data continuously through a wireless connection to a master node, typically a PDA or a PC, that collects, visualizes and analyzes the data. The current challenge in WBSN is the reduction of power consumption to reduce the dependency on batteries and allow miniaturization of the nodes. Reduction of power consumption can be done at the level of each component composing the sensor node, and the trade-off between local Digital Signal Processing (DSP) and radio transmission turns out to be very important. It has been shown that, with off-the-shelf radio such as ZigBee, it is always more attractive to process the data locally on the node, therefore reducing the amount of data to be transmitted, before sending it out through the radio link for reception by the master node. Consequently, research groups in WBSN have started investigating the possibility of implementing their algorithms for local on-node DSP.

The purpose of this project is to develop embedded signal processing for WBSN. More specifically, the project will start from four existing algorithms for ECG analysis and will aim at optimizing them for real-time signal processing and implementation on chip. The four algorithms that will be considered perform heart rate calculation and ECG signal delineation and exhibit variable computing complexity. The microprocessor targeted for this project is the TI MSP430, an 8 MHz processor chosen for its low-power characteristics. 8MHz are known to be sufficient to implement basic algorithms for heart rate calculation, whereas slightly more complex algorithm will require some code optimization. The C-codes corresponding to the four algorithms will be provided. The project will include the understanding of those codes at an algorithmic level, optimization for real-time and embedded implementation and the actual implementation on the microprocessor. Finally, sensor nodes will be provided such that the algorithms could be tested on-the-field with real ECG signals. The results of this project will serve as a first benchmark for DSP in WBSN research.

Tasks of the student:

For each of the four algorithms, the student will:

  • Get familiar with the C-codes and understand the algorithm behind them.
  • Modify and optimize the code for real-time implementation.
  • Modify and optimize the code for embedded implementation on a low-power microprocessor with limited computing power.
  • Port the algorithms onto the TI MSP430 microprocessor.
  • Test their implementation on a body sensor network platform.

To conclude, the student will derive a short comparison study for the four algorithms that will include, but not be limited to: algorithm complexity (number of cycles, operations), adequacy for real-time applications, adequacy for on-chip DSP and level of optimization required.

Note: if time allows, estimation of power consumption of some of the algorithms would be a plus in the project.


  • Advanced programming experience in C.
  • Good theoretical knowledge of embedded software requirements.
  • Basic practical experience with embedded software.
  • Interest in biomedical applications and practical work on wireless sensor nodes.


This project was supervised by Nadia Khaled.


***Archived: This is a past project,  it is not on offer at the moment!***