Accelerometer Technologies, Specifications, and Limitations: A presentation by ActiGraph at ICAMPAM 2013

Introduction
Technology advances and manufacturing efficiency improvements have drastically increased the options and flexibility available to users of accelerometer-based products. These advancements have led to confusion among activity monitor users, and many have been led to believe that device output normalization is an obvious and easy step. In truth, there are many engineering obstacles that make normalization extremely complicated.

Purpose
The purpose of this presentation is to educate users on the differences between integrated circuit accelerometer technologies and the tradeoffs and burdens that may be incurred by selecting one type over another, regardless of manufacturer. This paper will work to dispel the myth that accelerometers and/or activity monitors can easily be used interchangeably while furthering the understanding of accelerometers and their capabilities within the arena of human activity monitoring.

Summary
Until recent years, researchers interested in monitoring and quantifying human activity only had one viable option. The technology of Microelectromechanical systems (MEMS) was not yet cost effective for everyday use, and as a result, accelerometer-based activity monitors relied primarily on piezoelectric bimorph beams. These beams, while functional, are typically expensive to manufacture, require periodic calibration, and are limited to measurement of time varying acceleration, which precludes positional information such as subject posture. Advances in silicon wafer and manufacturing processing have enabled MEMS based accelerometers to become prevalent in many applications, including nearly every smartphone manufactured today. This technology is extremely stable, exhibits negligible measurement drift due to temperature, and requires only a single calibration. Furthermore, because they are capable of measuring static acceleration, positional information can be harvested for various applications.

These advances in the field of MEMS technology, coupled with the decreasing cost per bit for non-volatile memory, have led to a fundamental shift in the way activity data are collected. Researchers are no longer limited by the long standing filtered/epoch level data collection that dominated the arena for so many years. They are now free to collect raw acceleration data. This approach maximizes flexibility, allowing researchers to post process and reprocess data as new algorithms become available. With this additional flexibility comes a growing interest by the research community in normalizing outputs across devices, thereby removing the unique value provided by individual manufacturers. While this idea holds great promise among those tasked with harvesting useful information from the collected data, there are numerous hurdles that prevent it from being easily achieved. Proper education can establish appropriate data collection expectations, reduce confusion about the data collected, and allow for realistic comparisons between different accelerometer based products.

Download the full white paper here: ActiGraph ICAMPAM 2013