WAVELET DE-NOISING FOR AUTONOMOUS LATITUDE DETERMINATION

Abstract

Autonomous determination of the latitude of the place of movable and immovable objects is used as an independent task, as well as the task of determination of the initial value of latitude for operation of both platform and platform-free navigation systems. To solve these problems, it is necessary to have an inertial measurement unit (IMU) with at least three gyroscopes and three accelerometers. When using the IMU, executed by MEMS technology, the output signals of micromechanical gyroscope and accelerometers have significant noise components. Kalman filter is usually used to filter such signals. However, for this purpose it is necessary to know, besides the exact mathematical model of sensitive elements, many of their initial random characteristics.
 In the article, the research was conducted in order to investigate the use of wavelet transformation for the filtering of output signals of micromechanical accelerometers and gyroscopes for autonomous determination of the latitude of the place. The peculiarity of using wavelet transform for noisy signals is that due to changing scale, wavelets can detect differences in process characteristics on different scales, and with help of the shift we can analyze process properties at different points on the whole investigated interval. Due to the properties of this system's fullness that it is possible to restore the process by means of inverse wavelet transform. The efficiency of the developed method of increasing the accuracy of the autonomous determination of the latitude of the IMU on the basis of micromechanical gyroscope and accelerometers has been experimentally confirmed. The projections of the angular velocity of Earth rotation and gravitational acceleration were obtained from the IMU made by MEMS technology. After that, the signals of the gyroscopes and accelerometers of the inertial measuring unit were filtered, using the wavelet ‘Daubechies 10’ in decomposition, and averaged. These signals were used in a computational algorithm to determine the latitude. The results showed that, unlike the well-known Kalman filter, which almost did not increase the accuracy of the latitude calculation, wavelet denoising and further averaging reduced calculation error by almost twice.