Efficient Sensor Signal Filtering for Autonomous Wireless Nodes

Wireless sensor networks are nowadays a reality. However, a wireless node must be a truly autonomous system, i.e., it must embed its own power source, which is generally a battery pack. For this reason, it is of main concern to limit every source of wasted power. In particular, nodes exploit low-duty-cycle strategies to increase their life, spending most of their time in low-power mode, turning off all devices within the system except a wake-up oscillator. However, this way, some time must be spent waiting for the circuitry settling time, particularly if a high-accuracy measurement is required. In this paper, an efficient implementation of a filtering algorithm that is able to reduce this time is discussed. It is based on a hybrid combination of a median and a mean filter, joining the advantages of both linear and nonlinear filtering. A well-tailored implementation has experimentally been tested and discussed. Two wireless prototypes have been realized to test the proposed filter performance for temperature and humidity measurements. The experimental results permit comparing the implemented filters in terms of computational time, power consumption, and the performance of noise filtering. For both sensors, the new filtering strategy resulted to be more efficient in terms of noise removal and consumption than traditional algorithms. In particular, in the case of slow sensors such as a humidity one, delays are dictated by the sensor itself, and computational time can be neglected, while for fast sensors such as a temperature one, the proposed schema greatly improves the system update rate.