Accelerometer Calibration Method Based on Polynomial Curve Fitting

Abstract

Measurement is a process to find out the object quantity in a certain unit. Accelerometer sensor is an inertial measurement unit that can be used to measure the motion states of certain objects either static or dynamic. The accelerometer as a measurement tool must be reliable and valid in expressing the value. So, the accelerometer must be calibrated first before being used to measure the motion state of the object. In this paper, we propose the polynomial curve fitting method for calibrating the accelerometer sensor. Basically, this accelerometer sensor works based on the Analog to Digital Converter (ADC) principle where it converts the tilt of the sensor to the corresponding voltage. It should be noted that this accelerometer consists of a triple-axis where all of the axes have the same input-output value. Hence, by collecting the data that contains a number of the tilts of the sensor and the corresponding voltages, it is possible to generate the mathematical model that maps the tilts of the accelerometer sensor to the corresponding voltages. From the experiment, we can generate the five-order polynomials model that can be used to predict the new value that approximates the ground-truth value. It can be proved by measuring the Mean Absolute Error (MAE) score of the polynomial curve fitting between the ground-truth value and the prediction value. As a result, the Mean Absolute Error (MAE) score for each of the axes is 0.57. It indicates that our proposed method based on the polynomial curve fitting has been successfully applied for calibrating the accelerometer sensor.