| Streszczenie: | In healthcare technology, failure to accurately detect a medical event, such as electromyographic (EMG) activity or the onset of a heart attack, could have serious consequences for the user. The application of sensor networks to optimize event detection is therefore a key area of biomedical engineering research. The optimum number of sensors for event detection can be difficult and time-consuming to determine experimentally. A novel probabilistic model could be used to estimate the appropriate number of sensors required and is described in this paper. A simple statistically independent model (naïve approach) is introduced, which may be useful but relies on a set of assumptions that do not reflect most real-world applications. From there, a more practical, revised model that can be more easily applied to real-life systems was developed. Both models were subsequently tested against experimental data from a real EMG system. While the naïve approach was practically unachievable, the model was found to be mathematically sound. Predictions from the models were compared and the revised model was found to be more accurate than the naïve model. The revised model was verified against a majority voting system and it successfully predicted sensor network performance with only a small margin of error. This work presents a probabilistic sensor network approach to virtually exploring possible networks and informing optimum design of sensor networks used in healthcare and industry.
|
|---|