SYSTEM FOR DETERMINING THE COORDINATES OF THE ACOUSTIC SIGNAL SOURCE BASED ON THE DIFFERENCE-TIME METHOD AND MACHINE LEARNING METHODS
DOI:
https://doi.org/10.31891/2219-9365-2023-75-1Keywords:
determining the coordinates of the signal source, acoustic signal, artificial neural network, time difference, sound location methods, sound sensors, location accuracy, intelligent systems, supervised learning, machine learningAbstract
This scientific article proposes a system for determining the coordinates of an acoustic signal source using an artificial neural network. The system uses the differences in the registration times of the acoustic signal between the sensors as input data to train the neural network. The neural network uses the differences in registration times to determine the X and Y coordinates of the corresponding source.
The authors proposed software-mathematical models for the location of sensors, which are used for training and testing the neural network, as well as for evaluating the metrological characteristics of the system for determining the coordinates of the acoustic signal source. Software-mathematical models make it possible to configure such parameters as: the shape of the location of the sensors, the distance between them, their number, the parameters of the neural network (the number of hidden layers and neurons), and the volume of the dataset for training. The dataset includes such training pairs as the difference in the time of registration of the acoustic signal and the X and Y coordinates of the corresponding signal. Two datasets were used to test the neural network. The first was used both for training the neural network and for testing. The second was not used for training, but only for testing. This paper presents the structural diagram of the system for determining the coordinates of the acoustic signal source, as well as the block diagram of the neural network training algorithm used in this study.
The study of the absolute error of determining the X and Y coordinates of the acoustic signal source for specific system parameters was carried out. The smallest and largest distance between sensors and sources of acoustic signals was 300 m and 1800 m, respectively. Based on the results of the research, it was established that the average value of the absolute error in determining the Y coordinate did not exceed 1.2 m, and for the X coordinate - 0.5 m. The maximum absolute error value did not exceed 16 m for the Y coordinate and 4.5 m for the X coordinate.
