Target Detection and Recognition System and Method Based on Wireless Optical Communication

Abstract

A target detection and recognition system and method based on wireless optical communication is disclosed, comprising: transmitting coded and modulated optical signals; receiving optical signals reflected by a target object and converting them into electrical signals; preprocessing, demodulating, and decoding the received electrical signals; performing time series analysis and frequency domain analysis on the preprocessed signals to extract features of the target object; and classifying and recognizing the extracted features using machine learning methods. The system and method of this invention improve the accuracy and efficiency of target detection and recognition, and reduce reliance on complex hardware, making the system more economical and easier to deploy. In particular, the innovative solutions of this invention for optical signal processing, analysis, and recognition, such as the innovative design and efficient combination of signal feature extraction and deep learning processing methods, enhance its detection and recognition capabilities and application scope.

Description

Technical Field

[0001] This invention relates to the field of wireless optical communication technology, and in particular to a target detection and recognition system and method based on wireless optical communication. Background Technology

[0002] In today's increasingly automated and intelligent world, target detection and recognition technology plays a vital role and has been widely applied in various fields such as industrial manufacturing, safety inspection, and waste recycling. For example, in industrial manufacturing, it is used to inspect product quality; in safety inspection, it is used to identify potential hazardous substances; and in waste recycling, this technology can efficiently distinguish between different types of waste, effectively reducing environmental pollution. The efficient application of target detection and recognition technology can significantly improve operational efficiency and accuracy, especially in scenarios requiring rapid processing and analysis of large numbers of samples. Furthermore, this technology can also be applied in the field of health monitoring, such as fall detection for the elderly, to improve home safety.

[0003] Currently, target detection and recognition technology has evolved from simple shape and color recognition to complex multi-dimensional recognition of materials, textures, and shapes. This mainly relies on several technical means: X-ray and ultrasound technologies, while capable of penetrating deep into the interior of targets, are typically expensive, bulky, and pose potential health risks to operators. Camera-based vision methods, while intuitive and easy to implement, are often limited by ambient lighting and changing viewing angles, and may involve privacy issues. Radio frequency technology and Wireless Local Area Network (WLAN) technology, although widespread and low-cost, have limited recognition accuracy in complex environments.

[0004] In summary, existing technologies face challenges in terms of cost, portability, environmental adaptability, and accuracy.

[0005] It should be noted that the information disclosed in the background section above is only for understanding the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The main objective of this invention is to overcome the deficiencies of the aforementioned background technology and provide a target detection and recognition system and method based on wireless optical communication.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, a target detection and recognition system based on wireless optical communication includes: A wireless optical communication transmitter module, used to transmit encoded and modulated optical signals; A wireless optical communication receiver module is used to receive light signals reflected by a target object and convert them into electrical signals. The signal preprocessing unit is used to preprocess, demodulate, and decode the received electrical signals; The feature extraction unit is used to perform time series analysis and frequency domain analysis on the preprocessed signal to extract the features of the target object; The machine learning processing unit is used to classify and identify the extracted features using machine learning methods.

[0008] When executing the above modules or units in specific embodiments, they can be operated in the manner described below. It should be understood that the following specific methods used in the implementation of this invention are only illustrative examples and do not constitute a limitation on this invention. These specific methods are optional, and the order of operation and combination relationship can also be adjusted or replaced by other methods. The scope covered by this invention includes, but is not limited to, the following methods listed.

[0009] Optionally, but not limitingly, the feature extraction unit extracts the shape, texture, material, and state features of the target object by analyzing the amplitude variation, spectral characteristics, waveform patterns, and time series features of the signal; wherein, the feature extraction unit identifies important frequency components in the signal, including peak frequency and bandwidth, through frequency domain analysis; and the feature extraction unit analyzes the long-term and short-term variation patterns of the signal through time series analysis to identify the unique waveform features of the target object.

[0010] Optionally, but not limitingly, the signal preprocessing unit further segments and optimizes the sampling of the signal data using a sliding window method to generate multiple data segments; wherein the sliding window method includes setting a window length to match the time span of the target object features, and generating overlapping data segments by gradually moving the window at intervals smaller than the window length; the overlapping data segments are used to preserve the temporal context information of the signal and introduce redundancy to enhance the robustness of the model to noise and interference; wherein the window length and the moving interval are set according to the characteristics of the optical signal and the target object.

[0011] Optionally, but not limitingly, the machine learning method may include the K-nearest neighbor algorithm, the random forest algorithm, or the support vector machine algorithm.

[0012] Optionally, but not limitingly, the machine learning processing unit may be a deep learning processing submodule, which includes: A bidirectional input layer is used for parallel processing of forward and reverse time series data; A bidirectional BiLSTM layer, which contains two LSTM layers, one for forward processing and one for backward processing of time series data; Residual connections are provided between the two LSTM layers of the bidirectional BiLSTM layer to facilitate direct gradient backflow. Fully connected layers are used to enhance feature propagation and learn more complex data representations; Functional layers are used to adapt to the processing of multi-channel time series data and to transform multi-dimensional data into a three-dimensional structure suitable for LSTM layer processing; Two residual BiLSTM layers are used for deep temporal dependency analysis; The classification layer, which consists of two fully connected layers and a SoftMax function, is used to perform multi-class classification and converts the model's output into a probability distribution. The Dropout component, integrated into the LSTM network, avoids overfitting by randomly shutting down a portion of neurons during training.

[0013] In a second aspect, a target detection and recognition method based on wireless optical communication includes the following steps: Transmit encoded and modulated optical signals; It receives light signals reflected by the target object and converts them into electrical signals; The received electrical signals are preprocessed, demodulated, and decoded. Time series analysis and frequency domain analysis are performed on the preprocessed signal to extract the features of the target object; The extracted features are classified and identified using machine learning methods.

[0014] In a third aspect, an electronic device includes a memory, a communication module, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the method.

[0015] In some embodiments, the wireless optical communication transmitter module performs the functions of encoding, modulating, and transmitting optical signals. During the encoding stage, error detection and correction mechanisms are incorporated to improve signal stability and transmission quality. In the subsequent modulation stage, the module modulates the input encoded signal to a type suitable for transmission through the wireless optical communication channel, further ensuring efficient signal transmission and integrity. Finally, the transmitter converts the input digital signal into an optical signal, a process involving the conversion of an electrical signal into a light waveform that can be emitted by an optical transmitter or light-emitting device, such as a light-emitting diode (LED). The wireless optical communication receiver module receives, demodulates, and decodes the optical signals emitted by the transmitter. This module utilizes highly sensitive light-sensing devices, such as photodiodes, to accurately capture the optical signals emitted by the transmitter and convert them back into electrical signals. In the demodulation stage, the module demodulates the corresponding modulation information from these electrical signals, ensuring accurate signal recovery. The decoding process further processes these signals, performing error checking and correction, thereby guaranteeing the accuracy and integrity of data transmission.

[0016] In some embodiments, the processor module is responsible for receiving and extracting the target reflected light signal from the receiver. First, it performs signal preprocessing, including amplification, filtering, and noise reduction, to optimize signal quality and readability. Then, the processor employs feature extraction techniques, such as machine learning or deep learning algorithms, to perform detailed analysis of the preprocessed signal. Through these algorithms, the module can identify the unique features of various targets and extract detailed information about the target's material, shape, texture, and state from the signal. Furthermore, by accurately analyzing the reflected light signal, the processor module can enable fall detection and daily activity recognition in health monitoring applications.

[0017] In some embodiments, during signal preprocessing, filters can be used to initially filter the signal, eliminating irrelevant frequency components originating from ambient light or electronic devices as much as possible to reduce the impact of external interference on the signal; signal scaling and standardization are then performed to adjust the data to a suitable scale and range to adapt to different environmental conditions and improve the accuracy and efficiency of subsequent processing; to enhance the robustness of the model and ensure continuity, a sliding window method is used to segment and optimize sampling of the data. These steps together ensure that the signal transmitted from the wireless optical communication receiving module is clear and accurate, providing high-quality input data for the complex algorithm analysis of target detection and recognition.

[0018] In some embodiments, the feature extraction step extracts features related to the material, shape, texture, and state of the preprocessed signal. This process involves detailed analysis of the signal to identify specific patterns of various features. Through time series analysis and frequency domain analysis, key information that distinguishes different targets is extracted from the signal, including but not limited to changes in light signal intensity, frequency distribution, and waveform characteristics. These together constitute a comprehensive description of the target's material, shape, texture, and state, and provide necessary data for further classification and identification processes.

[0019] In some embodiments, machine learning or deep learning methods are used to classify and identify the extracted features. By learning the features of known targets, newly received signals can be effectively classified, thereby accurately determining the target's material, shape, texture, and state. Furthermore, the system possesses the ability to continuously learn and adapt; as the dataset expands, the accuracy and efficiency of its detection and identification can be continuously improved.

[0020] The present invention has the following beneficial effects: This invention provides a target detection and recognition system and method based on wireless optical communication. It effectively utilizes the technological advantages of wireless optical communication to achieve accurate detection and recognition of the material, shape, texture, and state of target objects. This invention uses wireless optical communication equipment to achieve non-contact, non-destructive, and passive target detection and recognition, reducing the hardware dependence of traditional recognition systems. The system has advantages such as accurate non-contact detection and recognition, simple structure, low cost and energy consumption while maintaining accuracy, ease of deployment, and compatibility with existing communication and network infrastructure.

[0021] The advantages of this invention are its simple structure, low cost, and ease of deployment. Its main advantages include non-contact, accurate detection and identification capabilities, making the system suitable for sensitive or fragile objects; strong environmental adaptability, enabling stable operation under different lighting conditions; high data transmission efficiency, suitable for real-time analysis; and enhanced security and suitability for privacy-protecting applications.

[0022] In particular, the combination of frequency domain analysis and time series analysis methods enables the system to effectively process complex signals. Frequency domain analysis focuses on identifying frequency components in the signal, such as peak frequency and bandwidth, which are helpful in capturing the periodicity and frequency characteristics of reflected signals. Time series analysis, on the other hand, focuses on the changes in the signal over time, which is helpful in identifying the unique waveform features of the target object. The combined application of these two analysis methods in this invention improves the system's recognition accuracy and robustness, and maintains high accuracy in changing operating environments.

[0023] In a preferred embodiment, the present invention segments and optimizes the sampling of signal data using a sliding window method. By setting an appropriate window length to match the time span of the target object features, and by gradually sliding the window at small intervals, overlapping data segments are generated, preserving the temporal context information of the signal, while introducing redundancy, thereby enhancing the robustness of the model to noise and interference. By designing a gradually sliding small-interval moving window, the characteristics of the optical signal and changes in the target object are adapted, further improving the quality of signal processing.

[0024] In the preferred embodiment, the deep learning processing submodule constructed by this invention, through its innovative structural design, not only improves the accuracy and efficiency of target detection and recognition but also enhances its adaptability to complex environments. The deep learning processing submodule of the preferred embodiment of this invention performs excellently in target detection and recognition tasks, providing highly accurate and reliable results in both static and dynamic environments. In particular, the bidirectional input layer and bidirectional BiLSTM layer in the deep learning processing submodule enable the model to process forward and reverse time series data in parallel, thereby more comprehensively understanding and predicting the behavioral patterns of target objects. The introduction of residual connections effectively solves the gradient vanishing problem in deep networks, allowing the construction of deeper network structures without sacrificing training efficiency and accuracy. Furthermore, the collaborative work of fully connected layers and functional layers enhances the model's ability to process high-dimensional data, enabling the system to extract rich feature information from multi-channel time series data. Simultaneously, the deep temporal dependency analysis of the two residual BiLSTM layers further improves the model's ability to capture long-term dependencies. Building on this, the integration of the Dropout component effectively avoids overfitting and improves the model's generalization ability by randomly shutting down a portion of neurons during training.

[0025] Furthermore, the system and method of the present invention have multi-band compatibility, that is, the wireless optical communication light source of the present invention is applicable to other spectral ranges such as visible light, infrared (near infrared, mid infrared, far infrared), ultraviolet, and terahertz bands. The corresponding wireless optical communication technologies also include communication technologies based on visible light, optical camera communication (OCC), infrared, ultraviolet, terahertz, etc., to adapt to a wider range of application scenarios and improve the system's flexibility and environmental adaptability.

[0026] Other beneficial effects of the embodiments of the present invention will be further described below. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the target detection and recognition system according to an embodiment of the present invention; Figure 2This is a flowchart of the target detection and recognition system according to an embodiment of the present invention; Figure 3 This is a simplified flowchart of the target detection and recognition method according to an embodiment of the present invention; Figure 4 This is a simplified diagram of the network structure of the deep learning processing submodule in an embodiment of the present invention.

[0028] Figure 5 This is a schematic diagram of the electronic device structure according to an embodiment of the present invention. Detailed Implementation

[0029] The embodiments of the present invention will be described in detail below. It should be emphasized that the following description is merely exemplary and not intended to limit the scope and application of the present invention.

[0030] See Figures 1 to 2 This invention provides a target detection and recognition system based on wireless optical communication, including a wireless optical communication transmitter module, a wireless optical communication receiver module, a signal preprocessing unit, a feature extraction unit, and a machine learning processing unit. The wireless optical communication transmitter module is used to transmit encoded and modulated optical signals; the wireless optical communication receiver module is used to receive optical signals reflected by a target object and convert them into electrical signals; the signal preprocessing unit is used to preprocess, demodulate, and decode the received electrical signals; the feature extraction unit is used to perform time series analysis and frequency domain analysis on the preprocessed signals to extract features of the target object; and the machine learning processing unit is used to classify and recognize the extracted features using machine learning methods.

[0031] The target detection and recognition system and method based on wireless optical communication provided in this invention can effectively utilize the technical advantages of wireless optical communication to achieve accurate detection and recognition of the material, shape, texture, and state of target objects. This invention utilizes wireless optical communication equipment to achieve non-contact, non-destructive, and passive target detection and recognition, reducing the hardware dependence of traditional recognition systems. The system has advantages such as accurate non-contact detection and recognition, simple structure, low cost while maintaining accuracy, ease of deployment, and compatibility with existing communication and network infrastructure.

[0032] When executing the above modules or units in specific embodiments, they can be operated in the manner described below. It should be understood that the following specific methods used in the implementation of this invention are only illustrative examples and do not constitute a limitation on this invention. These specific methods are optional, and the order of operation and combination relationship can also be adjusted or replaced by other methods. The scope covered by this invention includes, but is not limited to, the following methods listed.

[0033] In a preferred embodiment, the feature extraction unit extracts the shape, texture, material, and state features of the target object by analyzing the amplitude variation, spectral characteristics, waveform patterns, and time series features of the signal; wherein, the feature extraction unit identifies important frequency components in the signal, including peak frequency and bandwidth, through frequency domain analysis; and the feature extraction unit analyzes the long-term and short-term variation patterns of the signal through time series analysis to identify the unique waveform features of the target object.

[0034] By combining frequency domain analysis and time series analysis, the system and method of this invention can effectively process complex signals. Frequency domain analysis focuses on identifying frequency components in the signal, such as peak frequency and bandwidth, which are helpful in capturing the periodicity and frequency characteristics of reflected signals. Time series analysis, on the other hand, focuses on the changes in the signal over time, which is helpful in identifying the unique waveform features of the target object. The combined application of these two analysis methods in this invention improves the system's recognition accuracy and robustness, and maintains high accuracy in changing operating environments.

[0035] In a preferred embodiment, the signal preprocessing unit further segments and optimizes the sampling of signal data using a sliding window method to generate multiple data segments. The sliding window method includes setting a window length to match the time span of the target object's features, and generating overlapping data segments by gradually moving the window at intervals smaller than the window length. The overlapping data segments are used to preserve the temporal context information of the signal and introduce redundancy to enhance the model's robustness to noise and interference. The window length and moving interval are set according to the characteristics of the optical signal and the target object.

[0036] See Figure 4In a particularly preferred embodiment of the present invention, the machine learning processing unit employs a deep learning method. The machine learning processing unit is equipped with a deep learning processing submodule, which includes a bidirectional input layer, a bidirectional BiLSTM layer, residual connections, a fully connected layer, a functional layer (not shown), two residual BiLSTM layers, a classification layer (dense layer), and a Dropout component (not shown). The bidirectional input layer is used for parallel processing of forward and reverse time series data; the bidirectional BiLSTM layer contains two LSTM layers, which process the time series data in forward and reverse directions respectively; the residual connections are located in the BiLSTM layer... The STM layer facilitates direct gradient backflow between two LSTM layers; the fully connected layer enhances feature propagation and learns more complex data representations; the functional layer adapts to the processing of multi-channel time series data and transforms multidimensional data into a three-dimensional structure suitable for LSTM layer processing; two residual BiLSTM layers are used for deep temporal dependency analysis; the classification layer (dense layer) includes two fully connected layers and a SoftMax function to perform multi-class classification, transforming the model's output into a probability distribution; and the Dropout component, integrated into the LSTM network, avoids overfitting by randomly shutting down a subset of neurons during training.

[0037] The deep learning processing submodule constructed in this invention, through its innovative structural design, not only improves the accuracy and efficiency of target detection and recognition but also enhances its adaptability to complex environments. The deep learning processing submodule of the preferred embodiment of this invention performs exceptionally well in target detection and recognition tasks, providing highly accurate and reliable results in both static and dynamic environments. In particular, the bidirectional input layer and bidirectional BiLSTM layer in the deep learning processing submodule enable the model to process forward and reverse time series data in parallel, thereby gaining a more comprehensive understanding and prediction of the target object's behavioral patterns. The introduction of residual connections effectively solves the gradient vanishing problem in deep networks, allowing the construction of deeper network structures without sacrificing training efficiency and accuracy. Furthermore, the collaborative work of fully connected layers and functional layers enhances the model's ability to process high-dimensional data, enabling the system to extract rich feature information from multi-channel time series data. Simultaneously, the deep temporal dependency analysis of the two residual BiLSTM layers further improves the model's ability to capture long-term dependencies. Building upon this, the integration of the Dropout component, by randomly shutting down a portion of neurons during training, effectively avoids overfitting and improves the model's generalization ability.

[0038] The following describes specific embodiments of the present invention.

[0039] Specifically, embodiments of this invention can achieve the detection and identification of a target's material, shape, texture, and state through reflected light from the target surface via wireless optical communication. By performing detailed preprocessing and feature extraction on the received light signal, and then using a pre-trained machine learning or deep learning model to classify these features, this invention can achieve efficient real-time identification of different targets.

[0040] This invention provides a target detection and recognition system based on wireless optical communication, including a wireless optical communication receiver module for receiving and converting optical signals, and a signal processing module with signal preprocessing, feature extraction, and machine learning or deep learning processing functions. This system enables non-contact target detection and recognition, effectively combining the efficiency and portability of wireless optical communication technology with the accuracy of related algorithms, providing an innovative technical solution for target detection and recognition.

[0041] Figure 1 This diagram illustrates the structural block diagram of the target detection and recognition system according to an embodiment of the present invention. The diagram explains the system's workflow and the interrelationships between the various modules. The process begins at the communication signal transmitter 100, which is responsible for transmitting optical signals for wireless optical communication. Next, the communication signal receiver 200 captures these signals and transmits them to the target detection and recognition module 300. In this module, the optical signals are converted and analyzed to extract relevant key features of the target for classification, thereby identifying the material, shape, texture, and state of different targets.

[0042] Both the communication signal transmitter 100 and the communication signal receiver 200 employ specially designed wireless optical communication equipment. This equipment is a signal transmission and reception integrated platform specifically designed for wireless optical communication, combining hardware and software to provide an efficient and reliable communication method. The core processing unit typically uses a high-performance microprocessor or microcontroller, responsible for performing all data processing and control tasks, including signal modulation, demodulation, and error handling. To achieve optical signal transmission and reception, the equipment is equipped with high-efficiency LEDs and highly sensitive photoelectric sensors. The LEDs not only emit stable light signals but also adjust the intensity and waveform of the light according to different communication requirements. Simultaneously, the photoelectric sensors are responsible for accurately capturing these light signals and converting them into electrical signals for further processing.

[0043] The communication signal transmitter 100 primarily utilizes the transmitter function of a wireless optical communication device, with its core consisting of a set of high-brightness LEDs. These LEDs, through precise control and modulation, can emit light signals of specific intensity and frequency to transmit complex data information. Depending on the communication protocol and environmental requirements, the transmitter can adjust parameters such as frequency, waveform, and light intensity of the light signal to ensure stable transmission and a high recognition rate. To adapt to different recognition scenarios, the transmitter can also adjust the signal coverage and direction, and can use other lamps and light sources depending on the type of wireless optical communication, demonstrating high adaptability and flexibility.

[0044] The communication signal receiver 200 primarily utilizes the receiving function of wireless optical communication equipment, mainly composed of photoelectric sensors, such as highly sensitive photodiodes. These sensors possess high sensitivity and a wide dynamic range, enabling them to accurately receive optical signals under various lighting conditions. To improve the reception quality and processing efficiency of the sensor signals, the receiver may also include a signal conditioning unit and a control unit.

[0045] The signal conditioning unit optimizes the received signal, including amplifying weak signals, filtering noise, and adjusting the signal range to suit the needs of subsequent processing circuits or units. It is equipped with multi-channel transimpedance amplifiers to enhance the received weak electrical signal and effectively convert it into a usable voltage signal. These amplifiers have built-in microcurrent amplification capabilities, enabling them to handle various signal strengths. To adapt to different signal conditions, the amplifiers are designed with adjustable matching resistors for flexible adjustment of the amplification factor.

[0046] The control unit processes the amplified electrical signal through a built-in analog-to-digital converter, converting it into a digital signal for subsequent analysis.

[0047] The target detection and recognition module 300 is responsible for processing the information received by the wireless optical communication system and accurately detecting and recognizing target characteristics. This module first preprocesses the received signal, including signal amplification, noise filtering, and signal standardization, to ensure the efficiency and accuracy of subsequent processing. Then, the module analyzes the preprocessed signal through feature extraction, using techniques such as time series analysis and frequency domain analysis to extract key feature information. Based on this, the module uses machine learning or deep learning algorithms to classify and recognize these features to determine the target object's material, shape, texture, and state.

[0048] Figure 2This diagram illustrates a workflow of the target detection and recognition system according to an embodiment of the present invention. The diagram details the different components of the system and their functions: the communication transmitter is responsible for transmitting wireless optical communication signals, while the communication receiver captures optical signals reflected from different targets. Subsequently, the signals undergo preprocessing and classification analysis to extract the reflected optical signals from the targets. The diagram further illustrates the modulation, demodulation, encoding, and decoding processes of the optical signals, demonstrating the system's integrated sensing functionality.

[0049] Figure 3 This diagram illustrates a simplified flowchart of the target detection and recognition method according to an embodiment of the present invention. The target detection and recognition process includes signal reception, processing, feature extraction, and final detection and recognition. In step 301, a photoelectric sensor at the wireless optical communication receiver captures the communication signal reflected from the target surface. This signal contains feature information emitted from the transmitter and reflected back from the target surface. In step 302, the acquired communication signal undergoes signal processing. An algorithm identifies the presence of reflected signals, and then feature information characterizing the target's material, shape, texture, and state is extracted from these signals. Finally, in step 303, the target feature information obtained through processing and analysis is sent to a predetermined terminal device for further display, recording, and data analysis.

[0050] In this embodiment, the signal processing module includes signal preprocessing, feature extraction, and machine learning or deep learning processing sub-modules.

[0051] The signal preprocessing module first uses a low-pass filter (such as a Butterworth filter) to perform preliminary filtering of the signal to eliminate high-frequency components. This is based on the understanding of signal characteristics, which assumes that important information is usually contained in the low-frequency range, while the high-frequency components are likely external noise. Next, data scaling and standardization are performed to adapt to specific proportions or value ranges that may arise under different environments and communication light source intensities. The standardization process uses mean and variance standardization, with the specific transformation formula being... , where 𝜇 represents the mean and 𝜎 represents the standard deviation, ensures that all data are brought to a common scale, promotes more effective training, and ultimately improves prediction accuracy.

[0052] In a preferred embodiment, the signal preprocessing submodule uses a sliding window method to segment the data, with the window length adjusted to a suitable time span. By gradually sliding the window at small intervals, overlapping data segments are generated. This aims to preserve the temporal context of the recorded sequence and introduce some redundancy to enhance the robustness of the model. It also optimizes the balance between recognition accuracy and computational requirements, improving processing efficiency. During sampling, the data sequence is input into the model at an optimized length to balance recognition accuracy and computational requirements. The data points are downsampled to the predetermined sequence length using the system's sampling rate, and then feature extraction is performed on the data samples.

[0053] The feature extraction submodule processes relevant signals to extract features related to the target's material, shape, texture, and state, including signal amplitude variations, spectral characteristics, waveform patterns, and time-series features. Using frequency domain analysis, the module can identify important frequency components in the signal, such as peak values ​​and bandwidth, which are particularly important for distinguishing different targets. Simultaneously, through time-series analysis, the module analyzes the long-term and short-term variation patterns of the signal to identify unique waveform features specific to a particular target. These features constitute a comprehensive analysis of the target's characteristics, providing a foundation for subsequent identification.

[0054] The machine learning processing submodule can employ various machine learning algorithms, including K-Nearest Neighbors, Random Forest, and Support Vector Machines. The K-Nearest Neighbors algorithm classifies data points by analyzing similar data points in the signal feature space, while Random Forest and Support Vector Machines improve recognition accuracy by constructing multiple decision trees and selecting the optimal classification hyperplane, respectively.

[0055] See Figure 4In a preferred embodiment, the deep learning processing submodule is designed with the following deep learning network to process wireless optical communication signals for accurate identification. This module first integrates multiple Long Short-Term Memory (LSTM) layers, which are particularly well-suited for processing and predicting important event intervals and patterns in time series data because they can store past information and use it to influence the current output. These include bidirectional LSTM (BiLSTM) and BiLSTM with residual structures. Bidirectional LSTM improves the understanding of time series data through both forward and backward information flow, while the residual structure BiLSTM, which adds residual connections between BiLSTM layers, helps overcome the vanishing gradient problem common in deep networks, allowing for more direct gradient flow and enabling the construction of deeper models to capture more complex features. Next, several fully connected layers are stacked to enhance feature propagation; these layers enable the model to learn more complex data representations through non-linear transformations. The model includes a series of functional layers to adapt to the processing needs of multi-channel time series data, for example, by transforming multi-dimensional data into a three-dimensional structure suitable for LSTM. The model also includes a bidirectional input layer with 32 neurons and two residual BiLSTM layers for deep temporal dependency analysis. Furthermore, the dense layers utilize two fully connected layers and the SoftMax function for multi-class classification, where the SoftMax function transforms the model's output into a probability distribution for easier classification. To avoid overfitting and improve the network's generalization ability, the LSTM network incorporates a dropout component, which prevents the model from becoming overly reliant on specific patterns in the training data by randomly "turning off" neurons during training. The algorithm is implemented in Python using PyTorch and trained offline on a server equipped with a high-performance Graphics Processing Unit (GPU). The GPU accelerates the model's training process, making it more efficient.

[0056] Figure 5 A schematic diagram of an electronic device according to an embodiment of the present invention is shown. The electronic device includes a memory 401, a processor 402, and a communication module 403. The electronic device includes a computer program stored in the memory 401 and executable on the processor 402, implementing a target detection and recognition method based on wireless optical communication.

[0057] Memory 401 is the core storage component of the electronic device, responsible for storing all necessary running programs and data. It may include high-speed random access memory (RAM) and one or more non-volatile memories (NVM), such as solid-state drives or disk storage, to ensure persistent storage and fast access to data.

[0058] Processor 402, as the computing core of the electronic device, is responsible for executing computer programs stored in memory 401. This program contains instructions for processing and executing target detection and recognition methods based on wireless optical communication. The processor can be a Central Processing Unit (CPU), an Application Specific Integrated Circuit (ASIC), or other forms of processing unit, specifically optimized for high-speed, accurate data processing and target detection and recognition.

[0059] The communication module 403 is responsible for data communication between the processor 402 and the memory 401, as well as with external devices. It supports various communication protocols and bus standards, such as Industry Standard Architecture (ISA), Peripheral Component Interconnect (PCI), or Extended Industry Standard Architecture (EISA), ensuring efficient and reliable data transmission.

[0060] Optionally, in a specific implementation, the memory, processor, and communication interface can be integrated on the same chip and communicate through an internal interface, thereby improving the overall system integration and efficiency.

[0061] In summary, the main advantages of this invention compared to the prior art are: This invention proposes a target detection and recognition system and method based on wireless optical communication, which improves the accuracy and efficiency of recognition, reduces reliance on complex hardware, and makes the system more economical and easier to deploy. In particular, the innovative solutions of this invention in optical signal processing, analysis, and recognition, such as the innovative design and efficient combination of signal feature extraction and deep learning processing methods, enhance its detection and recognition capabilities and application scope.

[0062] This invention also provides a storage medium for storing a computer program, which, when executed, performs at least the methods described above.

[0063] The storage medium can be implemented by any type of non-volatile storage device, or a combination thereof. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), magnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disc, or compact disc read-only memory (CD-ROM); the magnetic surface memory can be a disk drive or magnetic tape drive. The storage media described in the embodiments of this invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0064] In the several embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0065] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0066] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0067] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0068] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0069] The methods disclosed in the several method embodiments provided by this invention can be arbitrarily combined without conflict to obtain new method embodiments.

[0070] The features disclosed in the several product embodiments provided by this invention can be arbitrarily combined without conflict to obtain new product embodiments.

[0071] The features disclosed in the several method or device embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new method or device embodiments.

[0072] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various equivalent substitutions or obvious modifications can be made without departing from the concept of the present invention, and all such modifications, achieving the same performance or application, should be considered within the scope of protection of the present invention.

Claims

1. A target detection and recognition system based on wireless optical communication, characterized in that, include: A wireless optical communication transmitter module, used to transmit encoded and modulated optical signals; A wireless optical communication receiver module is used to receive light signals reflected by a target object and convert them into electrical signals. The signal preprocessing unit is used to preprocess, demodulate, and decode the received electrical signals; The feature extraction unit is used to perform time series analysis and frequency domain analysis on the preprocessed signal to extract the features of the target object; The feature extraction unit extracts the shape, texture, material, and state features of the target object by analyzing the amplitude changes, spectral characteristics, waveform patterns, and time series features of the signal. Specifically, the feature extraction unit identifies important frequency components in the signal, including peak frequency and bandwidth, through frequency domain analysis. Furthermore, the feature extraction unit analyzes the long-term and short-term variation patterns of the signal through time series analysis to identify the unique waveform features of the target object. The machine learning processing unit is used to classify the extracted features using machine learning methods to identify the attributes of the target object, including the material, shape, texture, and state of the target object.

2. The target detection and recognition system as described in claim 1, characterized in that, The signal preprocessing unit further segments and optimizes the sampling of signal data using a sliding window method to generate multiple data segments. The sliding window method involves setting a window length to match the time span of the target object's features and moving the window at small, incremental intervals to generate overlapping data segments. These overlapping data segments are used to preserve the signal's temporal context information and introduce redundancy to enhance the model's robustness to noise and interference. The window length and moving interval are set according to the characteristics of the optical signal and the target object.

3. The target detection and recognition system as described in any one of claims 1 to 2, characterized in that, The machine learning method mentioned employs the K-nearest neighbor algorithm, random forest algorithm, or support vector machine algorithm.

4. The target detection and recognition system as described in any one of claims 1 to 2, characterized in that, The machine learning method described employs a deep learning approach based on a multi-layered neural network structure using LSTM.

5. A target detection and recognition method based on wireless optical communication, characterized in that, Includes the following steps: Transmit encoded and modulated optical signals; It receives light signals reflected by the target object and converts them into electrical signals; The received electrical signals are preprocessed, demodulated, and decoded. The preprocessed signal is subjected to time series analysis and frequency domain analysis to extract the features of the target object. Specifically, the amplitude variation, spectral characteristics, waveform patterns and time series features of the signal are analyzed to extract the shape, texture, material and state features of the target object. In particular, the important frequency components in the signal, including peak frequency and bandwidth, are identified through frequency domain analysis. Through time series analysis, the long-term and short-term variation patterns of the signal are analyzed to identify the unique waveform features of the target object. The extracted features are classified using machine learning methods to identify the attributes of the target object, including the object's material, shape, texture, and state.

6. The target detection and recognition method as described in claim 5, characterized in that, The signal data is further segmented and optimized using a sliding window method to generate multiple data segments. This sliding window method involves setting a window length to match the time span of the target object's features, and gradually moving the window at intervals smaller than the window length to generate overlapping data segments. These overlapping data segments are used to preserve the signal's temporal context information and introduce redundancy to enhance the model's robustness to noise and interference. The window length and moving interval are set according to the characteristics of the optical signal and the target object.

7. The target detection and recognition method as described in any one of claims 5 to 6, characterized in that, The machine learning method is a deep learning method, which classifies and identifies the extracted features through a deep learning processing network. The deep learning processing network is a multi-level neural network structure based on LSTM.

8. An electronic device comprising a memory, a communication module, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 5-7.

Images