Despite the relatively short period during which artificial neural networks have been used in system identification and control, there is already a rich history and a vast amount of literature describing successful applications [1,2]. Most reported achievements are in the areas of process control, robotics and manufacturing. There is no doubt that the use of such networks as been a major development in the field of control of nonlinear systems [2]. A newer area of applications for neural networks is that of instrumentation systems, more specifically, the development of intelligent transducers. As instrumentation equipment and measurement procedures become more automated, the need for sophisticated sensors and a control approach increases [3]. This is particularly true when developing acceleration measurement systems [4].Since their inception, the acceleration sensors have generally been complex electromechanical devices, consisting of relatively large proof masses, hinges and servos [5]. Recent advances in micro-electro-mechanical system (MEMS) technologies have made possible silicon acceleration sensors of very small size and with low power consumption [5]. Such features permit a wide range of possible applications where motion/movement-controlled systems are used [6]. However, in spite of the advances in micromachining, no sensor is perfect in its manufacture and the capacitive sensors considered here is no exception [3,4].These devices not only exhibit non-idealities such as offset, drift, non-linearity and noise, but also the magnitude of these non-idealities can vary. Moreover, fundamental characteristics of the sensor, e.g. sensitivity, may be subject to manufacturing tolerances, varying material properties and ambient effects [6]. Compensation of time-variant ambient effects requires continuous monitoring of these effects and on-line correction of the sensor behavior [3]. On the other hand, time-invariant departures from ideal behavior can be corrected using single-shot calibration procedures. Both correction procedures may require additional hardware and software and must therefore be considered during the design phase of the sensor system [5].In this paper, the issues of identification and control of such sensors are approached with the aim of developing improved performance transducers. Several neural networks based smart transducer designs are discussed, and their performance compared to that of open-loop, off-the-shelf capacitive acceleration sensors.