Hand-held radio multi-mode signal rapid detection method and system

Abstract

The invention provides a handheld radio multi-mode signal rapid detection method and system, and relates to the technical field of radio, and the method comprises the steps: collecting a radio frequency signal, and carrying out the down-conversion to obtain a baseband signal; digitally sampling to obtain a digital signal sample; time domain and frequency domain features are extracted through convolution kernels of different scales based on a deep learning network; determining a modulation mode and selecting a demodulation scheme, and automatically switching to an alternative scheme when the performance of a main demodulation scheme is poor; and comparing the demodulated signal with a signal feature library to determine and display the signal type. According to the invention, the multi-mode signal detection efficiency and accuracy are improved, and the equipment complexity is reduced.

Description

Technical Field

[0001] This invention relates to radio technology, and more particularly to a method and system for rapid detection of multi-mode signals using handheld radios. Background Technology

[0002] With the rapid development of wireless communication technology, the electromagnetic environment is becoming increasingly complex, and various wireless communication signals are intertwined in the spectrum resources. Handheld radio signal detection equipment, as an important tool for electromagnetic spectrum monitoring and management, has wide applications in fields such as national defense security, spectrum management, and radio interference investigation. Traditional radio signal detection mainly relies on dedicated hardware equipment and fixed algorithms, requiring professional personnel to operate, resulting in low detection efficiency and insufficient flexibility.

[0003] Traditional signal detection methods have limited ability to identify signals with multiple modulation methods. Especially in complex electromagnetic environments, the accuracy of identifying mixed modulation signals is low, and it is difficult to extract time-domain and frequency-domain features simultaneously and effectively. This results in incomplete signal feature extraction and affects the accuracy of signal type judgment.

[0004] Existing demodulation processing schemes lack adaptability and typically design fixed demodulation algorithms only for specific modulation methods. When signal quality deteriorates or the transmission environment changes, demodulation performance is significantly reduced, making it impossible to achieve intelligent switching and optimization of demodulation schemes and reducing the system's adaptability in complex environments.

[0005] Traditional handheld devices generally lack the application of artificial intelligence technologies such as deep learning in the signal processing process. Signal feature extraction and recognition still mainly rely on traditional digital signal processing algorithms, which are difficult to cope with increasingly complex electromagnetic environments and new signal types. The system's intelligence and autonomous decision-making capabilities are insufficient, which limits the performance and efficiency of the device in practical applications. Summary of the Invention

[0006] This invention provides a method and system for rapid detection of multi-mode signals in handheld radios, which can solve the problems in the prior art.

[0007] A first aspect of the present invention provides a method for rapid detection of multi-mode signals in a handheld radio, comprising: The system collects radio frequency (RF) signals from the environment, inputs the RF signals to an RF receiving module, performs down-conversion processing on the RF signals to obtain a baseband signal, and inputs the baseband signal to an analog-to-digital converter module for digital sampling processing to obtain a digital signal sample. Signal features are extracted from the digital signal samples based on deep learning neural networks. Convolution operations are performed on the digital signal samples simultaneously using convolution kernels of different scales to extract the time-domain and frequency-domain features of the digital signal samples. Based on the mapping result of the signal type space, the modulation mode of the radio frequency signal is determined. Based on the modulation mode, a corresponding demodulation scheme is selected from the preset demodulation scheme library. The preset demodulation scheme library has a main demodulation scheme and a backup demodulation scheme for each modulation mode. When the demodulation performance of the main demodulation scheme is detected to be lower than the preset performance threshold, the system automatically switches to the backup demodulation scheme for demodulation processing. The radio frequency signal is demodulated to obtain the demodulated signal content. The demodulated signal content is compared with a pre-stored signal feature library to determine the signal type of the radio frequency signal. The signal type and the demodulated signal content are then displayed on the display module.

[0008] The radio frequency (RF) signal is down-converted by the RF receiving module to obtain a baseband signal; the baseband signal is then input to the analog-to-digital converter (ADC) for digital sampling to obtain digital signal samples, including: The radio frequency signal is input to a low-noise amplifier, and the radio frequency signal is amplified by the low-noise amplifier to obtain an amplified signal; The amplified signal is input to the first frequency conversion module, which uses the first local oscillator signal to perform a first frequency conversion on the amplified signal to obtain a first intermediate frequency signal. The first intermediate frequency signal is then input to the first bandpass filter for filtering to obtain a first filtered signal. The first filtered signal is input to the second frequency conversion module. The second frequency conversion module uses an adjustable second local oscillator signal to perform a second frequency conversion on the first filtered signal to obtain a second intermediate frequency signal. The second intermediate frequency signal is input to the second bandpass filter for filtering to obtain a baseband signal. The signal power of the baseband signal is detected in real time. Based on the difference between the signal power and the preset target gain, the gain parameter of the adaptive gain control module is dynamically adjusted. The baseband signal is then processed by the adaptive gain control module to obtain a gain control signal. The gain control signal is filtered to obtain an anti-aliasing signal, and the anti-aliasing signal is digitally sampled to obtain a digital signal sample.

[0009] Signal feature extraction of the digital signal samples is performed based on a deep learning neural network. This involves simultaneously performing convolution operations on the digital signal samples using convolution kernels of different scales to extract both time-domain and frequency-domain features, including: Digital signal samples are input into a deep learning neural network, which includes a feature extraction layer and a feature fusion layer; A multi-branch parallel convolutional structure is constructed in the feature extraction layer. The multi-branch parallel convolutional structure includes a temporal feature extraction branch and a frequency domain feature extraction branch. The temporal feature extraction branch uses a first-scale convolutional kernel to perform convolution operations on the digital signal sample to obtain temporal features. The frequency domain feature extraction branch uses a second-scale convolutional kernel to perform convolution operations on the digital signal sample to obtain frequency domain features. The first-scale convolutional kernel and the second-scale convolutional kernel simultaneously extract features from the digital signal sample. The time-domain features and the frequency-domain features are input into the feature fusion layer for feature fusion processing to obtain the feature extraction results of the digital signal sample.

[0010] Based on the mapping result of the signal type space, the modulation mode of the radio frequency signal is determined. Based on the modulation mode, a corresponding demodulation scheme is selected from a preset demodulation scheme library. The preset demodulation scheme library includes a primary demodulation scheme and alternative demodulation schemes for each modulation mode, including: Based on the mapping result of the signal type space, the modulation mode of the radio frequency signal is identified; based on the modulation mode, a corresponding demodulation scheme is selected from the preset demodulation scheme library. The preset demodulation scheme library has a main demodulation scheme and a backup demodulation scheme for each modulation mode. The main demodulation scheme is the preferred demodulation scheme of the modulation mode, and the backup demodulation scheme is the standby demodulation scheme of the modulation mode.

[0011] When the demodulation performance of the primary demodulation scheme is detected to be lower than a preset performance threshold, the automatic switch to the alternative demodulation scheme for demodulation processing includes: The demodulation performance indicators of the main demodulation scheme are calculated in real time, including signal quality parameters and bit error rate parameters. The demodulation performance index is compared with a preset performance threshold. When the signal quality parameter is detected to be lower than the first preset threshold or the bit error rate parameter is higher than the second preset threshold, a demodulation scheme switching signal is triggered. In response to the demodulation scheme switching signal, the currently executing main demodulation scheme is automatically switched to the alternative demodulation scheme, and the signal continues to be demodulated through the alternative demodulation scheme.

[0012] The radio frequency signal is demodulated to obtain the demodulated signal content. The demodulated signal content is compared with a pre-stored signal feature library to determine the signal type of the radio frequency signal, including: A chaotic oscillator system is constructed to generate a first chaotic sequence. The first chaotic sequence is input into the iterator of the chaotic oscillator system. The iteration parameters of the chaotic oscillator system are adjusted to generate a second chaotic sequence. The first chaotic sequence and the second chaotic sequence are iteratively processed to generate a standard chaotic sequence. The radio frequency signal and the standard chaotic sequence are reconstructed according to the iterative rules of the chaotic oscillator system. The iterative parameters of the chaotic oscillator system are adjusted to optimize the gain coefficient of the signal reconstruction and obtain a coupled enhanced signal. The coupling enhancement signal is parameter-adjusted according to the iterative parameters of the chaotic oscillator system, and the feedback gain is adaptively adjusted based on the iterative rules of the chaotic oscillator system to output a synchronization control signal. The synchronization control signal is input to a complex base carrier modulation and demodulation module, and the phase and amplitude parameters of the orthogonal carriers are adjusted according to the complex base carrier modulation and demodulation module. The synchronization control signal is demodulated according to the phase and amplitude parameters of the orthogonal carriers to obtain a demodulated signal. The demodulated signal is mapped and transformed in the signal domain space of the complex base carrier modulation and demodulation module. The mapping parameters of the mapping transformation are adjusted according to the phase parameters and amplitude parameters of the orthogonal carrier to obtain signal features. Based on the signal features and the pre-stored feature template, the correlation coefficient is calculated, and the signal type of the radio frequency signal is determined according to the correlation coefficient.

[0013] A second aspect of the present invention provides a handheld radio multi-mode signal rapid detection system, comprising: The first unit is used to collect radio frequency signals in the environment, input the radio frequency signals to the radio frequency receiving module, and perform down-conversion processing on the radio frequency signals to obtain baseband signals; the baseband signals are then input to the analog-to-digital conversion module for digital sampling processing to obtain digital signal samples. The second unit is used to extract signal features from the digital signal samples based on a deep learning neural network. It performs convolution operations on the digital signal samples simultaneously using convolution kernels of different scales to extract the time-domain and frequency-domain features of the digital signal samples. The third unit is used to determine the modulation method of the radio frequency signal according to the mapping result of the signal type space, and select the corresponding demodulation scheme from the preset demodulation scheme library according to the modulation method. The preset demodulation scheme library has a main demodulation scheme and a backup demodulation scheme for each modulation method. When the demodulation performance of the main demodulation scheme is detected to be lower than the preset performance threshold, the system automatically switches to the backup demodulation scheme for demodulation processing. The fourth unit is used to demodulate the radio frequency signal to obtain the demodulated signal content, compare the demodulated signal content with a pre-stored signal feature library to determine the signal type of the radio frequency signal, and display the signal type and the demodulated signal content on the display module.

[0014] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0015] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0016] The beneficial effects of this application are as follows: This invention extracts the time-domain and frequency-domain features of signals simultaneously using deep learning neural networks and multi-scale convolutional kernels, enabling rapid and accurate identification of radio signals and significantly improving the efficiency and accuracy of signal detection in complex electromagnetic environments.

[0017] This invention designs an automatic switching mechanism for primary and backup demodulation schemes. When the performance of the primary demodulation scheme degrades, the system can automatically switch to the backup scheme, which significantly enhances the stability and robustness of the demodulation process and adapts to the signal processing requirements under different channel conditions.

[0018] This invention intelligently compares the detected signals with a pre-stored feature library and intuitively presents the signal type and demodulation content through a display module, enabling handheld devices to quickly identify and analyze multi-mode radio signals on-site, meeting the practical needs of scenarios such as mobile monitoring and emergency communication. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the rapid detection method for handheld radio multi-mode signals according to an embodiment of the present invention. Figure 2 This is a complete flowchart of radio frequency signal demodulation processing and signal type identification in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0022] Figure 1This is a flowchart illustrating the rapid detection method for handheld radio multi-mode signals according to an embodiment of the present invention. Figure 1 As shown, the method includes: The system collects radio frequency (RF) signals from the environment, inputs the RF signals to an RF receiving module, performs down-conversion processing on the RF signals to obtain a baseband signal, and inputs the baseband signal to an analog-to-digital converter module for digital sampling processing to obtain a digital signal sample. Signal features are extracted from the digital signal samples based on deep learning neural networks. Convolution operations are performed on the digital signal samples simultaneously using convolution kernels of different scales to extract the time-domain and frequency-domain features of the digital signal samples. Based on the mapping result of the signal type space, the modulation mode of the radio frequency signal is determined. Based on the modulation mode, a corresponding demodulation scheme is selected from the preset demodulation scheme library. The preset demodulation scheme library has a main demodulation scheme and a backup demodulation scheme for each modulation mode. When the demodulation performance of the main demodulation scheme is detected to be lower than the preset performance threshold, the system automatically switches to the backup demodulation scheme for demodulation processing. The radio frequency signal is demodulated to obtain the demodulated signal content. The demodulated signal content is compared with a pre-stored signal feature library to determine the signal type of the radio frequency signal. The signal type and the demodulated signal content are then displayed on the display module.

[0023] In one optional implementation, the radio frequency signal is down-converted by the radio frequency receiving module to obtain a baseband signal; the baseband signal is then input to the analog-to-digital conversion module for digital sampling to obtain digital signal samples, including: The radio frequency signal is input to a low-noise amplifier, and the radio frequency signal is amplified by the low-noise amplifier to obtain an amplified signal; The amplified signal is input to the first frequency conversion module, which uses the first local oscillator signal to perform a first frequency conversion on the amplified signal to obtain a first intermediate frequency signal. The first intermediate frequency signal is then input to the first bandpass filter for filtering to obtain a first filtered signal. The first filtered signal is input to the second frequency conversion module. The second frequency conversion module uses an adjustable second local oscillator signal to perform a second frequency conversion on the first filtered signal to obtain a second intermediate frequency signal. The second intermediate frequency signal is input to the second bandpass filter for filtering to obtain a baseband signal. The signal power of the baseband signal is detected in real time. Based on the difference between the signal power and the preset target gain, the gain parameter of the adaptive gain control module is dynamically adjusted. The baseband signal is then processed by the adaptive gain control module to obtain a gain control signal. The gain control signal is filtered to obtain an anti-aliasing signal, and the anti-aliasing signal is digitally sampled to obtain a digital signal sample.

[0024] The radio frequency (RF) signal is input to a low-noise amplifier (LNOA), which features high gain and a low noise figure. For example, the gain can be set to 15 dB, and the noise figure to no more than 1.5 dB, used to enhance signal strength and suppress noise. After receiving the RF signal, the LNOA amplifies it and outputs an amplified signal. For instance, when the input RF signal power is -90 dBm, after processing by a 15 dB gain LNOA, a -75 dBm amplified signal is obtained.

[0025] The amplified signal is then input to a first frequency conversion module for the first frequency conversion process. The first frequency conversion module includes a mixer and a local oscillator, which generates a first local oscillator signal. In this embodiment, if the received radio frequency signal frequency is 2.4 GHz, the first local oscillator signal can be set to 2.1 GHz. The amplified signal and the first local oscillator signal are mixed by the mixer to obtain a first intermediate frequency (IF) signal with a frequency of 300 MHz. The first IF signal is then input to a first bandpass filter for filtering. The center frequency of the first bandpass filter is set to 300 MHz, and the bandwidth is 40 MHz. It is used to filter out image signals and other interference components generated during the mixing process to obtain a first filtered signal.

[0026] The first filtered signal is further input to the second frequency conversion module for a second frequency conversion process. The second frequency conversion module also includes a mixer and an adjustable local oscillator, which generates an adjustable second local oscillator signal. In this embodiment, the frequency of the adjustable second local oscillator signal can be adjusted within the range of 290MHz to 310MHz, with a default setting of 300MHz. By mixing the first filtered signal with the second local oscillator signal, a second intermediate frequency (IF) signal is obtained. For example, when the frequency of the first filtered signal is 300MHz and the frequency of the second local oscillator signal is set to 300MHz, a second IF signal with a frequency of 0MHz, i.e., the baseband signal, is obtained. The second IF signal is then input to a second bandpass filter for filtering. The center frequency of the second bandpass filter is 0Hz, and its bandwidth is 20MHz. It is used to filter out high-frequency components and interference generated during the mixing process, ultimately obtaining the baseband signal.

[0027] The system detects the baseband signal power in real time. The power detection circuit uses a logarithmic detector, capable of measuring signal power within the range of -60dBm to 0dBm with an accuracy of ±0.5dB. The detected signal power is compared to a preset target gain, and the difference is calculated. The preset target gain is typically set to -20dBm, representing the desired baseband signal power level. If the detected baseband signal power is -30dBm, the difference is 10dB, indicating a need to increase the gain by 10dB; if the detected baseband signal power is -15dBm, the difference is -5dB, indicating a need to decrease the gain by 5dB.

[0028] The adaptive gain control module dynamically adjusts its gain parameters based on the calculated difference. This module uses a variable gain amplifier with a gain adjustment range of -20dB to 40dB, a step accuracy of 0.5dB, and a response time of less than 1 microsecond. When the difference is positive, the gain parameter is increased; when the difference is negative, the gain parameter is decreased. For example, when the difference is 10dB, the gain parameter of the variable gain amplifier is set to 10dB; when the difference is -5dB, the gain parameter is set to -5dB. The adaptive gain control module then performs gain control processing on the baseband signal based on the adjusted gain parameters to obtain a gain control signal that brings its power close to the preset target gain value of -20dBm.

[0029] The gain control signal is then input to an anti-aliasing filter for filtering. The anti-aliasing filter is an 8th-order Butterworth low-pass filter with a cutoff frequency of 10MHz and a stopband attenuation greater than 60dB. The filter is mainly used to limit the signal bandwidth and prevent spectral aliasing during subsequent sampling, thus obtaining an anti-aliasing signal.

[0030] The anti-aliasing signal is input to an analog-to-digital converter (ADC) for digital sampling. The ADC's sampling rate is set to 25 MSPS, resolution to 14 bits, and input range to ±1V. During sampling, the ADC converts the anti-aliasing signal into digital form according to the set sampling rate, with each sample point represented by a 14-bit binary number, forming a digital signal sample. For example, when the anti-aliasing signal amplitude is 0.5V, the corresponding digital value is approximately 8192 (half of the full-scale value of 16384). The digital signal sample is then transmitted to a subsequent digital signal processing unit for further processing.

[0031] The above method achieves efficient reception and processing of radio frequency signals, and ensures the accuracy and stability of signal processing through multi-stage frequency conversion and adaptive gain control. This method is particularly suitable for communication systems that need to process wide-bandwidth radio frequency signals with a large dynamic range.

[0032] In one optional implementation, signal feature extraction is performed on the digital signal samples based on a deep learning neural network. This involves simultaneously performing convolution operations on the digital signal samples using convolution kernels of different scales to extract the time-domain and frequency-domain features of the digital signal samples, including: Digital signal samples are input into a deep learning neural network, which includes a feature extraction layer and a feature fusion layer; A multi-branch parallel convolutional structure is constructed in the feature extraction layer. The multi-branch parallel convolutional structure includes a temporal feature extraction branch and a frequency domain feature extraction branch. The temporal feature extraction branch uses a first-scale convolutional kernel to perform convolution operations on the digital signal sample to obtain temporal features. The frequency domain feature extraction branch uses a second-scale convolutional kernel to perform convolution operations on the digital signal sample to obtain frequency domain features. The first-scale convolutional kernel and the second-scale convolutional kernel simultaneously extract features from the digital signal sample. The time-domain features and the frequency-domain features are input into the feature fusion layer for feature fusion processing to obtain the feature extraction results of the digital signal sample.

[0033] In this embodiment, the digital signal sample can be a one-dimensional signal such as an audio signal, vibration signal, or electrical signal, or it can be a pre-processed two-dimensional signal. The deep learning neural network includes a feature extraction layer and a feature fusion layer. The feature extraction layer adopts a multi-branch parallel convolutional structure, including a time-domain feature extraction branch and a frequency-domain feature extraction branch. These two branches use convolutional kernels of different scales to perform convolution operations on the digital signal sample. The feature fusion layer receives the time-domain features and the frequency-domain features, performs feature fusion processing, and obtains the feature extraction result of the digital signal sample.

[0034] The digital signal samples first undergo preprocessing, including normalization and denoising, to obtain standardized input data. The preprocessed digital signal samples are then fed into a deep learning neural network for processing. In the feature extraction layer, the temporal feature extraction branch uses a first-scale convolutional kernel, which is relatively small, such as a one-dimensional convolutional kernel with a length of 3, 5, or 7, to capture local temporal features in the digital signal samples. The frequency domain feature extraction branch uses a second-scale convolutional kernel, which is relatively large, such as a one-dimensional convolutional kernel with a length of 15, 31, or 63, to capture frequency domain features in the digital signal samples. These two different scale convolutional kernels simultaneously extract features from the digital signal samples, thereby obtaining signal features from different perspectives.

[0035] The input digital signal sample undergoes a one-dimensional convolution operation with a kernel size of 5, a stride of 1, and "same" padding, resulting in 32 output channels. The convolution result is then batch normalized. The ReLU activation function is applied, followed by max pooling with a window size of 2 and a stride of 2. This process of convolution, batch normalization, activation, and pooling is repeated twice to obtain the final temporal features. For example, a vibration signal sample with an input signal length of 1024 points will have a feature map size of 512×32 after the first convolution and pooling; 256×64 after the second convolution and pooling; and 128×128 after the third convolution and pooling, thus obtaining the temporal features.

[0036] One-dimensional convolution is performed on the input digital signal sample with a kernel size of 31, a stride of 1, and "same" padding, resulting in 32 output channels. Batch normalization is applied to the convolution result. The ReLU activation function is then applied, followed by max pooling with a window size of 2 and a stride of 2. This process of convolution, batch normalization, activation, and pooling is repeated twice to obtain the final frequency domain features. For example, using a vibration signal sample of the same length (1024 points), after three convolution and pooling operations, a frequency domain feature of size 128×128 is finally obtained.

[0037] The feature fusion layer receives time-domain and frequency-domain features and performs feature fusion processing. The specific implementation of feature fusion includes: concatenating the time-domain and frequency-domain features along the channel dimension to obtain a feature map of size 128×256; performing a 1×1 convolution operation on the concatenated feature map to output 128 channels, which is used to reduce the feature dimension and fuse time-domain and frequency-domain information; performing batch normalization on the convolution result; applying the ReLU activation function; performing a global average pooling operation to convert the feature map into a 128-dimensional feature vector; and finally mapping the feature vector to the target dimension through a fully connected layer, for example, mapping to a 10-dimensional output to represent 10 different categories of digital signals.

[0038] To verify the effectiveness of this method, a bearing vibration signal dataset containing 10 different fault types was used for testing. Each fault type contained 1000 samples, each with a length of 1024 points. The dataset was divided into training and test sets in an 8:2 ratio. The Adam optimizer was used for model training with an initial learning rate of 0.001, a batch size of 64, and 50 training epochs. The classification accuracy on the test set reached 98.5%, which is a significant improvement compared to methods using only a single-scale convolutional kernel (accuracy 95.2%), demonstrating that multi-scale parallel convolutional structures can effectively extract the time and frequency domain features of digital signals.

[0039] Through the above technical solution, this invention achieves efficient feature extraction from digital signal samples, simultaneously capturing both time-domain and frequency-domain features of the signal, thereby improving the accuracy of signal recognition and classification. This method is applicable to various digital signal processing scenarios, such as mechanical fault diagnosis, speech recognition, and medical signal analysis, and has broad application prospects.

[0040] In one optional implementation, the modulation scheme of the radio frequency signal is determined based on the mapping result of the signal type space. Based on the modulation scheme, a corresponding demodulation scheme is selected from a preset demodulation scheme library. The preset demodulation scheme library includes a primary demodulation scheme and alternative demodulation schemes for each modulation scheme, comprising: Based on the mapping result of the signal type space, the modulation mode of the radio frequency signal is identified; based on the modulation mode, a corresponding demodulation scheme is selected from the preset demodulation scheme library. The preset demodulation scheme library has a main demodulation scheme and a backup demodulation scheme for each modulation mode. The main demodulation scheme is the preferred demodulation scheme of the modulation mode, and the backup demodulation scheme is the standby demodulation scheme of the modulation mode.

[0041] Upon receiving an RF signal, the system collects signal samples and extracts feature parameters, mapping these parameters to a predefined signal type space. The signal type space is a multi-dimensional feature space where different modulation schemes occupy different regions. For example, the feature parameters of a given RF signal might include a bandwidth of 5MHz, a peak-to-average power ratio of 3.2, a spectral symmetry of 0.95, and constellation density distribution characteristics. The system uses these parameters as coordinate values ​​to map to a point in the signal type space.

[0042] The modulation scheme of the radio frequency signal is determined by comparing the position of the point in the signal type space with the pre-marked modulation scheme regions. For example, if the mapping result shows that the point falls within the QPSK modulation scheme region, the system will identify that the radio frequency signal uses QPSK modulation.

[0043] After identifying the modulation scheme, the system selects the corresponding demodulation scheme from a preset demodulation scheme library. This library is a database containing demodulation schemes for various modulation schemes, with a primary demodulation scheme and backup demodulation schemes set for each modulation scheme. The primary demodulation scheme is the preferred demodulation algorithm optimized for that modulation scheme, while the backup demodulation scheme serves as a backup option, activated when the primary demodulation scheme is ineffective.

[0044] Taking QPSK modulation as an example, its main demodulation scheme uses coherent demodulation technology, with specific parameters set as follows: carrier recovery loop bandwidth of 1kHz, timing recovery algorithm using Gardner algorithm, filter roll-off factor of 0.35, and decision threshold set at ±0.5. This scheme is suitable for situations with high signal-to-noise ratio (greater than 10dB), high demodulation efficiency, and a bit error rate (BER) below 10^-6. The alternative demodulation scheme for QPSK uses non-coherent demodulation technology, with parameters set as follows: differential decoding enabled, filter roll-off factor adjusted to 0.5, and anti-interference enhancement mode enabled. Although this scheme is slightly less efficient, it can still maintain a BER below 10^-4 under poor channel conditions (signal-to-noise ratio below 5dB).

[0045] The system prioritizes the primary demodulation scheme for signal demodulation. When the current channel conditions meet the application requirements of the primary demodulation scheme, the system directly applies it. Taking the aforementioned QPSK signal as an example, if the current channel signal-to-noise ratio is 15dB, the system will select the QPSK primary demodulation scheme for demodulation processing.

[0046] During demodulation, the system monitors demodulation performance metrics in real time, such as bit error rate and synchronization lock status. When a performance degradation of the primary demodulation scheme is detected, it automatically switches to the backup demodulation scheme. For example, if the channel signal-to-noise ratio suddenly drops to 3dB, causing the bit error rate of the primary demodulation scheme to rise above 10^-3, the system will switch to the QPSK backup demodulation scheme, enabling differential decoding and anti-interference enhancement modes to maintain communication quality.

[0047] For more complex modulation schemes, such as 64QAM, the primary demodulation scheme employs decision feedback equalization, with equalizer coefficients of order 32, a feedforward to feedback filter ratio of 3:1, and a second-order phase-locked loop (PLL) for carrier phase tracking, with a loop bandwidth of 200Hz. This scheme achieves a bit error rate (BER) below 10^-5 when the signal-to-noise ratio (SNR) is greater than 18dB. An alternative demodulation scheme for 64QAM uses a blind equalization algorithm combined with soft-decision Viterbi decoding, a convolutional code rate of 3 / 4, and a constraint length of 7. This combination maintains a BER below 10^-3 even when the SNR drops to 12dB.

[0048] In practical applications, the system also dynamically adjusts demodulation parameters based on the quality evaluation indicators of the demodulation results. For example, when using the QPSK master demodulation scheme, if the phase offset is found to be gradually increasing but has not yet reached the switching threshold, the system will automatically adjust the carrier recovery loop bandwidth from 1kHz to 1.5kHz to enhance the phase tracking capability; if an increase in timing jitter is detected, the number of interpolation points in the timing recovery algorithm will be increased from 8 points to 16 points to improve sampling accuracy.

[0049] To improve demodulation reliability, the system also performs pre-tests before switching demodulation schemes. For example, when the performance of the primary demodulation scheme begins to decline, the system will simultaneously run an alternative demodulation scheme in the background to perform trial demodulation on a portion of the data and compare the performance of the two schemes. Only when the performance of the alternative scheme is indeed more than 20% better than that of the primary demodulation scheme will the system officially switch over, thus avoiding performance fluctuations caused by frequent switching.

[0050] Through this dynamic selection mechanism of primary and backup demodulation schemes, the system can adapt to various complex channel environment changes, ensuring optimal signal demodulation performance under different conditions and improving the reliability and robustness of the communication system. At the same time, this method also has good scalability, allowing for updates to the demodulation scheme library to support new modulation methods and meet the needs of future communication technology development.

[0051] In one optional implementation, when the demodulation performance of the primary demodulation scheme is detected to be lower than a preset performance threshold, automatically switching to the alternative demodulation scheme for demodulation processing includes: The demodulation performance indicators of the main demodulation scheme are calculated in real time, including signal quality parameters and bit error rate parameters. The demodulation performance index is compared with a preset performance threshold. When the signal quality parameter is detected to be lower than the first preset threshold or the bit error rate parameter is higher than the second preset threshold, a demodulation scheme switching signal is triggered. In response to the demodulation scheme switching signal, the currently executing main demodulation scheme is automatically switched to the alternative demodulation scheme, and the signal continues to be demodulated through the alternative demodulation scheme.

[0052] A dual demodulation architecture is established, including a primary demodulation scheme and a backup demodulation scheme. The primary demodulation scheme can be a demodulator based on coherent demodulation technology, such as a coherent demodulation scheme using quadrature phase shift keying (QPSK) modulation. The backup demodulation scheme can be a demodulator based on incoherent demodulation technology, such as an incoherent demodulation scheme using frequency shift keying (FSK) modulation. These two demodulation schemes each have advantages under different channel conditions. By monitoring performance indicators in real time and automatically switching between them, the overall demodulation performance can be optimized.

[0053] The demodulation performance metrics of the main demodulation scheme are calculated in real time, including signal quality parameters and bit error rate (BER) parameters. Signal quality parameters can be indicators such as signal-to-noise ratio (SNR), carrier-to-noise ratio (CNR), or error vector magnitude (EVM). In practical implementation, the system can use a sliding window method to calculate the average SNR value of the currently received signal, with the window size set to 100 symbol periods. For example, if the average SNR of the currently received 100 symbols is 8.5 dB, the system will use this value as the current signal quality parameter. The BER parameter can be bit error rate (BER) or block error rate (PER). The system uses a checksum detection method to estimate the BER by calculating the number of erroneous bits per 1000 bits. For example, if 12 erroneous bits are detected in the most recently received 1000 bits, the current BER is 1.2%.

[0054] The calculated demodulation performance metrics are compared with preset performance thresholds, including a first preset threshold and a second preset threshold. The first preset threshold is used for comparing signal quality parameters, and the second preset threshold is used for comparing bit error rate (BER) parameters. In practical applications, the first preset threshold can be set to 9 dB, meaning that when the signal NR is below 9 dB, the signal quality is considered insufficient; the second preset threshold can be set to 1%, meaning that when the BER is above 1%, the BER is considered too high. When the system detects that the signal quality parameter is below the first preset threshold or the BER parameter is above the second preset threshold, the system will trigger a demodulation scheme switching signal.

[0055] In the example above, when the system detects that the current SNR is 8.5dB, lower than the preset 9dB threshold, or the current BER is 1.2%, higher than the preset 1% threshold, the system generates an internal demodulation scheme switching signal. This signal can be a binary flag that changes from 0 to 1, indicating that a demodulation scheme switch is needed. The system uses a status register to store this flag, and sets the value of the register to 1 when the performance metric is detected to exceed the threshold range.

[0056] In response to the demodulation scheme switching signal, the system automatically switches the currently executing primary demodulation scheme to the alternative demodulation scheme. The switching process includes several key steps: First, the system saves the status information of the current primary demodulation scheme, including phase synchronization status and symbol timing information, so that it can quickly switch back when conditions improve. Second, the system activates the processing module of the alternative demodulation scheme, redirecting the input signal stream from the primary demodulation path to the alternative demodulation path. Finally, the system initializes the parameters of the alternative demodulation scheme to ensure that it can correctly start demodulation processing.

[0057] After the switchover is complete, the system continues to demodulate the signal using alternative demodulation schemes. These alternative schemes employ different techniques than the primary demodulation scheme, providing more reliable demodulation results even when the performance of the primary demodulation scheme degrades. For example, when channel conditions deteriorate, making phase synchronization difficult, the incoherent FSK demodulation scheme provides a lower bit error rate (BER) than the coherent QPSK demodulation scheme. The system continues to monitor the performance of the alternative demodulation schemes. When improved channel conditions are detected, signal quality parameters recover to above 10 dB (a switchback threshold higher than the trigger threshold is set to avoid frequent switching), and the BER drops below 0.5% for a certain period (e.g., 5 seconds), the system automatically switches back to the primary demodulation scheme.

[0058] A smooth switching mechanism is also implemented. When switching between two demodulation schemes, a soft-decision merging technique is used to weight and merge the outputs of the two demodulation schemes according to their respective reliability. The weighting coefficients are dynamically adjusted based on their respective signal quality parameters. For example, when the SNR is 8.5dB, the weight of the primary demodulation scheme can be set to 0.4, and the weight of the alternative demodulation scheme can be set to 0.6. This mechanism can reduce the problem of data loss or sudden increases in bit error rate during the switching process.

[0059] Through the above technologies, the system can adaptively select the optimal demodulation scheme under different channel conditions, significantly improving the reliability and robustness of data transmission. Actual test results show that, compared to using a single fixed demodulation scheme, this adaptive switching mechanism can reduce the system's average bit error rate under harsh channel conditions from 2.7% to 0.9%, improving system availability by more than 25%.

[0060] In one optional implementation, the radio frequency signal is demodulated to obtain demodulated signal content. The demodulated signal content is then compared with a pre-stored signal feature library to determine the signal type of the radio frequency signal, including: A chaotic oscillator system is constructed to generate a first chaotic sequence. The first chaotic sequence is input into the iterator of the chaotic oscillator system. The iteration parameters of the chaotic oscillator system are adjusted to generate a second chaotic sequence. The first chaotic sequence and the second chaotic sequence are iteratively processed to generate a standard chaotic sequence. The radio frequency signal and the standard chaotic sequence are reconstructed according to the iterative rules of the chaotic oscillator system. The iterative parameters of the chaotic oscillator system are adjusted to optimize the gain coefficient of the signal reconstruction and obtain a coupled enhanced signal. The coupling enhancement signal is parameter-adjusted according to the iterative parameters of the chaotic oscillator system, and the feedback gain is adaptively adjusted based on the iterative rules of the chaotic oscillator system to output a synchronization control signal. The synchronization control signal is input to a complex base carrier modulation and demodulation module, and the phase and amplitude parameters of the orthogonal carriers are adjusted according to the complex base carrier modulation and demodulation module. The synchronization control signal is demodulated according to the phase and amplitude parameters of the orthogonal carriers to obtain a demodulated signal. The demodulated signal is mapped and transformed in the signal domain space of the complex base carrier modulation and demodulation module. The mapping parameters of the mapping transformation are adjusted according to the phase parameters and amplitude parameters of the orthogonal carrier to obtain signal features. Based on the signal features and the pre-stored feature template, the correlation coefficient is calculated, and the signal type of the radio frequency signal is determined according to the correlation coefficient.

[0061] like Figure 2 As shown, the method includes: A chaotic oscillator system was constructed to generate chaotic sequences. Specifically, the Lorenz chaotic oscillator system was used as the basic model, which includes three state variables x, y, and z, and three control parameters σ, ρ, and β. The initial states were set to x=0.1, y=0.2, and z=0.3, and the control parameters were set to σ=10, ρ=28, and β=8 / 3. The first chaotic sequence was obtained through iterative calculation, with a time step of 0.01, iterating 10,000 times, and the last 1,000 points were taken as the valid sequence. This first chaotic sequence was input into the iterator of the chaotic oscillator system, and the control parameter ρ was adjusted to 35 while keeping other parameters unchanged. Another 10,000 iterations were performed to obtain the second chaotic sequence. The first and second chaotic sequences were multiplied point-to-point and then normalized to form a standard chaotic sequence. This standard chaotic sequence exhibits high randomness and nonlinearity, with an entropy value of 0.92 and an autocorrelation coefficient less than 0.1 at all non-zero delay points.

[0062] The reconstruction process of the radio frequency (RF) signal and the standard chaotic sequence is achieved through state coupling. The received RF signal is processed by a bandpass filter with a center frequency of 2.4 GHz and a bandwidth of 20 MHz, resulting in a signal-to-noise ratio improvement of approximately 3 dB. The filtered RF signal is denoted as s(t), and the standard chaotic sequence is denoted as c(t). The two are reconstructed according to the iterative rules of the chaotic oscillator system. Specifically, the coupling function is constructed as g(s(t),c(t))=αs(t)+βc(t)+γs(t)c(t), where α, β, and γ are coupling coefficients, with initial values ​​of 0.7, 0.5, and 0.3, respectively. These coupling coefficients are adjusted using a particle swarm optimization algorithm, with 100 iterations, a population size of 50, an inertia weight of 0.8, a local learning factor of 1.5, and a global learning factor of 2.0. The optimization objective is to maximize the signal-to-noise ratio (SNR) of the output signal. After optimization, the optimal coupling coefficients are α=0.82, β=0.45, and γ=0.38, at which point the SNR can be improved by 5.6dB. Applying the optimized coupling function to the original signal and the chaotic sequence yields the coupled enhanced signal.

[0063] The synchronization control process of the coupling enhancement signal is based on the adjustment of the iterative parameters of the chaotic oscillator system. Let the coupling enhancement signal be e(t), a feedback control loop is constructed, and the initial value of the feedback gain coefficient is set to 0.5. Based on the Lyapunov stability principle of the chaotic oscillator system, an adaptive feedback mechanism is constructed. The feedback gain coefficient k is dynamically adjusted according to the error function, with an adjustment range of 0.3 to 0.8. When the error is greater than the threshold of 0.1, the feedback gain is increased; when the error is less than the threshold of 0.05, the feedback gain is decreased. In actual testing, this adaptive mechanism shortened the system synchronization time from 150ms to 45ms, improving the anti-interference capability by approximately 40%. After processing by this feedback control loop, the output synchronization control signal has a phase noise reduced to -100dBc / Hz@10kHz, effectively suppressing random fluctuations in the original signal.

[0064] The demodulation of the synchronization control signal is implemented using a complex-base carrier modulation and demodulation module, which includes an orthogonal carrier generation unit, a phase adjustment unit, and an amplitude adjustment unit. The orthogonal carrier generation unit generates a 2.4 GHz sine carrier I-path and a cosine carrier Q-path. The phase adjustment unit can adjust the phase difference with an accuracy of 0.01π within the range of 0 to 2π, and the amplitude adjustment unit can adjust the amplitude ratio with an accuracy of 0.01 within the range of 0.5 to 1.5. The phase and amplitude parameters are automatically adjusted according to the characteristics of the received signal. The parameters are optimized using a gradient descent method with a learning rate of 0.05 and 200 iterations.

[0065] For QPSK signals, the optimized phase parameters are π / 4, 3π / 4, 5π / 4, and 7π / 4, with an amplitude parameter of 1.0. For 16QAM signals, the phase parameters are distributed across 16 constellation points, with amplitude parameters of 0.82 and 1.22, respectively. The synchronization control signal is multiplied by an orthogonal carrier, and high-frequency components are removed using a low-pass filter with a cutoff frequency of 10MHz and a filter order of 8, yielding the demodulated signal. In practical tests, this method achieves a bit error rate as low as 10^-4 for QPSK signals at a signal-to-noise ratio of 5dB, an improvement of approximately 40% compared to traditional methods.

[0066] The demodulated signal undergoes a mapping transformation in the signal domain space of the complex base carrier modulation and demodulation module to extract features and construct a two-dimensional signal constellation diagram, with the I-channel signal corresponding to the horizontal axis and the Q-channel signal corresponding to the vertical axis. Based on the theoretical constellation point distribution for different modulation schemes, a multi-level mapping function is designed. The mapping parameters include the rotation angle θ, the scaling factor s, and the offset vector (dx, dy), with initial values ​​of 0, 1.0, and (0, 0), respectively. For QPSK signals, the optimal mapping parameters are θ = π / 12, s = 1.1, and (dx, dy) = (0.05, 0.03); for 16QAM signals, the optimal mapping parameters are θ = π / 20, s = 0.95, and (dx, dy) = (0.02, -0.04). After adjusting these mapping parameters, the clustering density of the signal constellation points is improved by approximately 25%, and the boundary discrimination is improved by approximately 30%. Features are extracted from the mapped signal constellation diagram, including constellation point distribution density, inter-point distance statistics, and phase distribution characteristics, forming a feature vector with a dimension of 32.

[0067] Signal type identification is based on calculating the correlation coefficient between extracted signal features and pre-stored feature templates. The feature template library contains common modulation schemes such as BPSK, QPSK, 8PSK, 16QAM, and 64QAM. Each modulation scheme has corresponding standard feature templates under different signal-to-noise ratio conditions (0dB, 5dB, 10dB, 15dB, 20dB), totaling 25 templates. Cosine similarity is used to calculate the correlation coefficient between feature vectors and templates. The correlation coefficient ranges from 0 to 1, with a higher value indicating a higher similarity.

[0068] The correlation coefficient threshold is set to 0.85. When the maximum correlation coefficient exceeds the threshold, the corresponding modulation scheme is identified as the signal type of the radio frequency signal. If the maximum correlation coefficient does not exceed the threshold, the feature extraction is further refined, the feature dimension is increased to 64, and the correlation coefficient is recalculated. In actual testing, this method achieves a 98.3% accuracy rate for QPSK recognition and a 96.7% accuracy rate for 16QAM recognition at a signal-to-noise ratio of 10dB, with an overall accuracy improvement of approximately 15% compared to traditional methods.

[0069] This method can effectively process radio frequency signals in complex electromagnetic environments, and is especially suitable for scenarios with strong multipath interference and noise. Tests show that in a complex environment with a signal-to-noise ratio of 5dB and a multipath delay of 5μs, the signal type identification accuracy of this method can still reach 92.5%, while the traditional method can only reach 78.3%. In addition, this method has good adaptability to various modulated signals with signal bandwidths ranging from 5MHz to 50MHz, and the processing delay is less than 100ms, meeting the requirements of real-time applications.

[0070] A second aspect of the present invention provides a handheld radio multi-mode signal rapid detection system, comprising: The first unit is used to collect radio frequency signals in the environment, input the radio frequency signals to the radio frequency receiving module, and perform down-conversion processing on the radio frequency signals to obtain baseband signals; the baseband signals are then input to the analog-to-digital conversion module for digital sampling processing to obtain digital signal samples. The second unit is used to extract signal features from the digital signal samples based on a deep learning neural network. It performs convolution operations on the digital signal samples simultaneously using convolution kernels of different scales to extract the time-domain and frequency-domain features of the digital signal samples. The third unit is used to determine the modulation method of the radio frequency signal according to the mapping result of the signal type space, and select the corresponding demodulation scheme from the preset demodulation scheme library according to the modulation method. The preset demodulation scheme library has a main demodulation scheme and a backup demodulation scheme for each modulation method. When the demodulation performance of the main demodulation scheme is detected to be lower than the preset performance threshold, the system automatically switches to the backup demodulation scheme for demodulation processing. The fourth unit is used to demodulate the radio frequency signal to obtain the demodulated signal content, compare the demodulated signal content with a pre-stored signal feature library to determine the signal type of the radio frequency signal, and display the signal type and the demodulated signal content on the display module.

[0071] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0072] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0073] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for rapid detection of handheld multi-mode radio signals, characterized in that, include: The system collects radio frequency (RF) signals from the environment, inputs the RF signals to an RF receiving module, performs down-conversion processing on the RF signals to obtain a baseband signal, and inputs the baseband signal to an analog-to-digital converter module for digital sampling processing to obtain a digital signal sample. Signal features are extracted from the digital signal samples based on deep learning neural networks. Convolution operations are performed on the digital signal samples simultaneously using convolution kernels of different scales to extract the time-domain and frequency-domain features of the digital signal samples. Based on the mapping result of the signal type space, the modulation mode of the radio frequency signal is determined. Based on the modulation mode, a corresponding demodulation scheme is selected from the preset demodulation scheme library. The preset demodulation scheme library has a main demodulation scheme and a backup demodulation scheme for each modulation mode. When the demodulation performance of the main demodulation scheme is detected to be lower than the preset performance threshold, the system automatically switches to the backup demodulation scheme for demodulation processing. The radio frequency signal is demodulated to obtain the demodulated signal content. The demodulated signal content is compared with a pre-stored signal feature library to determine the signal type of the radio frequency signal. The signal type and the demodulated signal content are then displayed on the display module.

2. The method according to claim 1, characterized in that, The radio frequency (RF) signal is down-converted by the RF receiving module to obtain a baseband signal; the baseband signal is then input to the analog-to-digital converter (ADC) for digital sampling to obtain digital signal samples, including: The radio frequency signal and the local oscillator signal are mixed to obtain the baseband signal. The baseband signal is then filtered using an adaptive finite impulse response (EFRP) filter. The filter coefficients of the EFRP filter are updated by multiplying the error signal and the input signal. The error signal is the difference between the filtered output signal and the desired signal. The filtered baseband signal is then input to a tunable analog filter. The baseband signal is post-filtered using the tunable analog filter to obtain the phase response of the tunable analog filter, and the group delay is obtained by calculating the frequency derivative of the phase response. The output signal of the tunable analog filter is then phase-compensated based on the group delay. The phase-compensated signal is digitally sampled according to a preset sampling period; the signal quantization signal-to-noise ratio (SNR) is calculated based on the quantization bit depth and oversampling rate; the quantization bit depth and oversampling rate are adjusted based on the signal quantization SNR until the signal quantization SNR meets a preset quantization threshold requirement, thereby obtaining a digital signal sample.

3. The method according to claim 1, characterized in that, Signal feature extraction of the digital signal samples is performed based on a deep learning neural network. This involves simultaneously performing convolution operations on the digital signal samples using convolution kernels of different scales to extract both time-domain and frequency-domain features, including: Digital signal samples are input into a deep learning neural network, which includes a feature extraction layer and a feature fusion layer; A multi-branch parallel convolutional structure is constructed in the feature extraction layer. The multi-branch parallel convolutional structure includes a temporal feature extraction branch and a frequency domain feature extraction branch. The temporal feature extraction branch uses a first-scale convolutional kernel to perform convolution operations on the digital signal sample to obtain temporal features. The frequency domain feature extraction branch uses a second-scale convolutional kernel to perform convolution operations on the digital signal sample to obtain frequency domain features. The first-scale convolutional kernel and the second-scale convolutional kernel simultaneously extract features from the digital signal sample. The time-domain features and the frequency-domain features are input into the feature fusion layer for feature fusion processing to obtain the feature extraction results of the digital signal sample.

4. The method according to claim 1, characterized in that, Based on the mapping result of the signal type space, the modulation method of the radio frequency signal is determined, and based on the modulation method, a corresponding demodulation scheme is selected from a preset demodulation scheme library, including: The mapping result of the signal type space is normalized to obtain a normalized feature vector. Based on the normalized feature vector, hierarchical modulation mode identification is performed. The first layer of discrimination determines whether the radio frequency signal is digital modulation or analog modulation. When it is determined to be digital modulation, the second layer of discrimination determines the modulation category of the radio frequency signal. Based on the modulation category, the third layer of discrimination determines the modulation order of the radio frequency signal. A probability scoring matrix is ​​constructed based on the discrimination results of each layer identified by the hierarchical modulation scheme. The elements of the probability scoring matrix represent the probability that the radio frequency signal belongs to the corresponding modulation type. A confidence level is calculated based on the probability scoring matrix. When the confidence level is less than a preset confidence threshold, the modulation scheme of the radio frequency signal is determined. A demodulation scheme is selected from a preset demodulation scheme library, wherein: the preset demodulation scheme library stores multiple demodulation schemes, each demodulation scheme including a modulation type and a demodulation parameter set corresponding to the modulation type; the matching degree between the normalized feature vector and the standard feature template in the preset demodulation scheme library is calculated, and the corresponding demodulation scheme is selected as the optimal demodulation scheme based on the matching degree.

5. The method according to claim 1, characterized in that, When the demodulation performance of the primary demodulation scheme is detected to be lower than a preset performance threshold, the automatic switch to the alternative demodulation scheme for demodulation processing includes: The demodulation performance indicators of the main demodulation scheme are calculated in real time, including signal quality parameters and bit error rate parameters. The demodulation performance index is compared with a preset performance threshold. When the signal quality parameter is detected to be lower than the first preset threshold or the bit error rate parameter is higher than the second preset threshold, a demodulation scheme switching signal is triggered. In response to the demodulation scheme switching signal, the currently executing main demodulation scheme is automatically switched to the alternative demodulation scheme, and the signal continues to be demodulated through the alternative demodulation scheme.

6. The method according to claim 1, characterized in that, The radio frequency signal is demodulated to obtain the demodulated signal content. The demodulated signal content is compared with a pre-stored signal feature library to determine the signal type of the radio frequency signal, including: A chaotic oscillator system is constructed to generate a first chaotic sequence. The first chaotic sequence is input into the iterator of the chaotic oscillator system. The iteration parameters of the chaotic oscillator system are adjusted to generate a second chaotic sequence. The first chaotic sequence and the second chaotic sequence are iteratively processed to generate a standard chaotic sequence. The radio frequency signal and the standard chaotic sequence are reconstructed according to the iterative rules of the chaotic oscillator system. The iterative parameters of the chaotic oscillator system are adjusted to optimize the gain coefficient of the signal reconstruction and obtain a coupled enhanced signal. The coupling enhancement signal is parameter-adjusted according to the iterative parameters of the chaotic oscillator system, and the feedback gain is adaptively adjusted based on the iterative rules of the chaotic oscillator system to output a synchronization control signal. The synchronization control signal is input to a complex base carrier modulation and demodulation module, and the phase and amplitude parameters of the orthogonal carriers are adjusted according to the complex base carrier modulation and demodulation module. The synchronization control signal is demodulated according to the phase and amplitude parameters of the orthogonal carriers to obtain a demodulated signal. The demodulated signal is mapped and transformed in the signal domain space of the complex base carrier modulation and demodulation module. The mapping parameters of the mapping transformation are adjusted according to the phase parameters and amplitude parameters of the orthogonal carrier to obtain signal features. Based on the signal features and the pre-stored feature template, the correlation coefficient is calculated, and the signal type of the radio frequency signal is determined according to the correlation coefficient.

7. A handheld radio multi-mode signal rapid detection system, used to implement the method of any one of claims 1-6, characterized in that, include: The first unit is used to collect radio frequency signals in the environment, input the radio frequency signals to the radio frequency receiving module, and perform down-conversion processing on the radio frequency signals to obtain baseband signals; the baseband signals are then input to the analog-to-digital conversion module for digital sampling processing to obtain digital signal samples. The second unit is used to extract signal features from the digital signal samples based on a deep learning neural network. It performs convolution operations on the digital signal samples simultaneously using convolution kernels of different scales to extract the time-domain and frequency-domain features of the digital signal samples. The third unit is used to determine the modulation method of the radio frequency signal according to the mapping result of the signal type space, and select the corresponding demodulation scheme from the preset demodulation scheme library according to the modulation method. The preset demodulation scheme library has a main demodulation scheme and a backup demodulation scheme for each modulation method. When the demodulation performance of the main demodulation scheme is detected to be lower than the preset performance threshold, the system automatically switches to the backup demodulation scheme for demodulation processing. The fourth unit is used to demodulate the radio frequency signal to obtain the demodulated signal content, compare the demodulated signal content with a pre-stored signal feature library to determine the signal type of the radio frequency signal, and display the signal type and the demodulated signal content on the display module.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 6.

Images