Work

A New Dynamic Approach to the Classification of Moving Objects by Changes in Frequency Intensity in the Signal Spectr

Year: 2026
Type: article
Abstract: Context. Modern moving object detection systems have different application specifics, technical equipment and are designed for different scenarios of interaction with the detected object. The corresponding features determine the use of very different technical means and information processing processes. The proposed dynamic approach to the recognition of moving objects, based on the change in the intensities of frequencies in the signal spectrum, is focused on the choice of methods for classifying objects by discrete convolution or recurrent neural network according to the available memory resources, computing power, speed and accuracy requirements for the surveillance system as a whole. The specified approach avoided redundancy in the required equipment, leaving the ability to integrate with existing systems for generating control effects on the identified object.Objective. The purpose of the work is to propose a dynamic approach to moving object recognition based on changing frequency intensities in the signal spectrum, which provides the necessary accuracy, speed, and memory usage by using post-processing of information using discrete convolution methods or a recurrent neural network.Method. The proposed approach to classifying moving objects includes preliminary filtering of the received signal in the frequency domain for the purpose of spectral analysis and detection of characteristic features. At the classification cascade, where the identification of object types is carried out, depending on the specified criteria of cost, accuracy and speed of the entire observation system, a classification tool is selected, which can be a discrete convolution or a recurrent neural network. The conclusions generated using the proposed approach regarding the type of object are transferred to the execution logic block, which forms control actions or makes decisions according to the classification results.Results. The proposed functional scheme for processing information about a moving object performs the conversion of the reflected analogy signal into a digital packet with subsequent computer processing for visualization and decision-making regarding further actions of the analogy Doppler signal system before its digitization by the STM32 microcontroller and subsequent data transmission via the UART–USB interface for computer processing. The use of the fast Fourier transform module provides obtaining the frequency spectrum of the signal for further classification of the moving object and generation of appropriate control effects on it.Conclusions. The scientific novelty of the work lies in the creation of a new dynamic approach to the recognition of moving objects, which is based on the analysis of changes in the intensities of the frequency components of the signal spectrum. The proposed approach involves the use of signal post-processing methods based on discrete convolution or recurrent neural networks, which provides increased accuracy and speed of the recognition process while simultaneously reducing the amount of required computing resources and memory. The practical significance of the results obtained lies in the creation and implementation of a functional motion recognition system based on a Doppler sensor, which provides conversion of an analogy reflected signal into a digital data packet with subsequent computer processing for its visualization and formation of control effects on the detected object. A mathematical description of the object recognition process using the discrete Fourier transform for a function of a real variable has been developed, which provided the possibility of implementing the signal processing algorithm directly on a microcontroller
Language: English
Open Access status: diamond
OpenAlex ID: W7214930331
FWCI:
Cites: 23
Cited by:
Related to: 10