A radio communication system and method with intercom functionality

Abstract

The present application relates to the field of radio communication technology, in particular to a radio communication system with intercom function, comprising an ambient voice acquisition module, an induced electric signal core processing module and a signal amplification optimization module, wherein: the ambient voice acquisition module converts ambient voice into induced electric signal; the induced electric signal core processing module defines the induced electric signal with the same sampling interval as the induced time sequence signal; judges the noise electric signal and the current analog electric signal in the induced time sequence signal; sets multiple tap numbers and corresponding tap coefficients, and sequentially adopts different tap numbers to reduce the noise of the current analog electric signal to output the filtered signal; sets the constraint condition to minimize the error signal, analyzes the optimal tap number and the corresponding tap coefficient; the signal amplification optimization module receives the optimal tap number in the induced electric signal core processing module, dynamically sets different voice feature amplification multiples according to the optimal tap number, and amplifies the current analog electric signal.

Description

Technical Field

[0001] This invention relates to the field of radio communication technology, and more specifically, to a radio communication system and method with intercom functionality. Background Technology

[0002] In the field of modern communications, wireless communication systems with intercom functionality are key communication tools in scenarios such as emergency dispatch, industrial production, and public safety. They typically employ microphones containing diaphragms, coils, and permanent magnets to collect ambient voice signals. These signals are then converted into induced electrical signals, which are subsequently processed using noise reduction techniques to generate a radio frequency (RF) hybrid signal (establishing a mapping relationship between the amplified signal and the high-frequency carrier amplitude, causing the high-frequency carrier amplitude to change synchronously with the amplified signal amplitude). This RF hybrid signal is then transmitted into space via an antenna. The receiving end, after capturing the RF hybrid signal through a receiver and antenna, extracts the current analog electrical signal containing voice information based on a preset amplitude mapping relationship, thus completing the wireless communication transmission of the voice signal and realizing the intercom function.

[0003] However, in practical applications, there are technical bottlenecks that make it difficult to meet the requirements for high-quality and high-reliability voice transmission. On the one hand, when the system terminal microphone (including diaphragm, ultra-fine copper wire coil and neodymium iron boron strong magnet) converts ambient voice (sound waves) into induced electrical signals, due to its working principle (diaphragm drives coil to cut magnetic field lines to generate induced electromotive force) and structural limitations, the initial signal amplitude is extremely small (millivolt level or even microvolt level), which is easily affected by electromagnetic interference, circuit characteristics and sound propagation characteristics, resulting in noise mixing, which leads to a decrease in signal clarity and intelligibility, and in severe cases, hinders the accurate transmission of information.

[0004] Existing technologies often employ methods such as moving average filtering, Wiener filtering, wavelet threshold denoising, adaptive filtering, and deep learning-based filtering to reduce noise in induced electrical signals, avoiding signal quality degradation and impacting intercom performance caused by noise. In this radio communication system, to address the dynamic fluctuations in noise within the scene (such as changes in electromagnetic interference with equipment startup / shutdown / personnel movement, and circuit thermal noise fluctuating with temperature), and considering that traditional fixed-parameter filtering (such as moving average filtering) has fixed parameters and is prone to problems such as "incomplete noise suppression" or "loss of voice details," an adaptive filtering method is typically adopted: setting the number of filter taps and initial tap coefficients, inputting a noisy induced electrical signal (a mixture of desired signal and noise signal) and a reference noise signal, calculating the error signal between the output and the desired signal using algorithms such as the least mean square algorithm, and then adjusting the tap coefficients in real time based on the error to minimize the error, ultimately outputting a filtered signal that retains the useful signal to the greatest extent and suppresses noise;

[0005] However, existing adaptive filters often use fixed or empirical values ​​when setting the number of taps. This is because induced electrical signals in radio communication scenarios are dynamic and complex—containing both steady-state speech (such as long vowels, with complex spectra requiring a sufficient number of taps to match the harmonic structure) and transient speech (such as voiceless consonants, which change rapidly and require an appropriate number of taps to avoid losing details). Furthermore, the intensity and frequency characteristics of mixed noise (such as electromagnetic noise and circuit thermal noise) change with the scenario (e.g., electromagnetic interference fluctuations caused by equipment start-up and shutdown, and thermal noise changes caused by temperature variations). If the number of taps is too high, the filter is prone to overfitting random noise, resulting in residual false noise after noise reduction and damage to signal purity. If the number of taps is too low, the filter cannot fully distinguish the frequency components of the signal, making it difficult to preserve the harmonic structure of steady-state speech and easily losing the plosive characteristics of transient speech. Ultimately, the noise reduction effect is limited because it cannot balance noise suppression and useful signal integrity. In view of this, we propose a radio communication system and communication method with intercom functionality. Summary of the Invention

[0006] The purpose of this invention is to solve the problem that noise interference during radio communication leads to a decrease in signal quality and poor communication effect, which in turn affects the stable and clear implementation of the intercom function.

[0007] To achieve the above objectives, the present invention provides a wireless communication system with intercom functionality, comprising an environmental voice acquisition module, an inductive electrical signal core processing module, and a signal amplification and optimization module, wherein:

[0008] The environmental speech acquisition module uses a microphone to acquire environmental speech and converts it into an induced electrical signal. The induced electrical signal core processing module receives the induced electrical signals acquired at t time points within the same sampling interval and defines them as induced time-series signals. It judges the noise electrical signal and the current analog electrical signal in the induced time-series signal. It sets multiple tap numbers and corresponding tap coefficients, and sequentially uses different tap numbers to denoise the current analog electrical signal and outputs a filtered signal. It obtains the desired signal by Fourier transforming the noise electrical signal and the current analog electrical signal in the induced time-series signal, calculates the error signal, sets constraints to minimize the error signal, iterates the tap coefficients and tap numbers, and determines the optimal tap number and corresponding tap coefficients. The signal amplification and optimization module receives the optimal tap number from the induced electrical signal core processing module, dynamically sets different speech feature amplification factors based on the optimal tap number, and amplifies the filtered signal.

[0009] As a further improvement to this technical solution, the microphone in the environmental voice acquisition module includes a diaphragm, and a coil is fixed on the back of the diaphragm; the coil is suspended in a constant magnetic field formed by a permanent magnet, and the axial direction of the coil is perpendicular to the direction of the magnetic field lines.

[0010] After the microphone receives the drive signal, the ambient voice propagates to the microphone in the form of sound waves. The sound waves of the ambient voice act on the diaphragm, and the diaphragm generates a periodic pressure difference and reciprocating mechanical vibration. When the diaphragm vibrates, it synchronously drives the coil to reciprocate linearly with the sound waves. The coil continuously cuts the magnetic field lines generated by the permanent magnet. According to the law of electromagnetic induction, an induced electromotive force is generated at both ends of the coil, which is an induced electrical signal that is consistent with the changing pattern of the ambient voice.

[0011] As a further improvement to this technical solution, the core processing module for the induced electrical signal includes a signal frame energy discrimination unit and an optimal tap number analysis and noise reduction unit. The signal frame energy discrimination unit receives the induced timing signal collected at t times in the same sampling interval, divides the induced timing signal into frames according to the frame length N and the frame shift R, calculates the short-time energy of each frame, and determines whether the induced timing signal of each frame is the current analog electrical signal. If it is not the current analog electrical signal, it is judged to be a noise electrical signal.

[0012] The optimal tap number analysis and noise reduction unit uses different numbers of taps and corresponding tap coefficients to filter the current analog electrical signal and output a filtered signal. It also extracts multiple delayed analog electrical signals corresponding to different tap numbers, assigns tap coefficients to the delayed analog electrical signals, and outputs a filtered signal by weighted summation. The unit calculates and estimates the noise electrical signal using the current analog electrical signal and the noise electrical signal, and calculates the desired signal. Based on the desired signal, it iterates the tap coefficients and the number of taps to select the optimal number of taps and tap coefficients.

[0013] As a further improvement to this technical solution, the signal frame energy discrimination unit sums the squares of the amplitudes of all sampling points of the induced timing signal in each frame after framing to obtain the short-time energy of the induced timing signal in each frame; an energy threshold is set, and if the energy threshold is less than the short-time energy, the corresponding frame is judged to be a noisy electrical signal; if the energy threshold is greater than or equal to the short-time energy, the corresponding frame is judged to be the current analog electrical signal, and then the current analog electrical signal is filtered and denoised by the optimal tap number analysis and denoising unit.

[0014] The beneficial effect of the above-mentioned further scheme is that by summing the squares of the amplitudes of all sampling points of the induction timing signal in each frame after framing, the short-time energy is obtained and then compared with the set energy threshold, so as to accurately distinguish the noise electrical signal and the current analog electrical signal, providing a basis for subsequent noise reduction filtering processing only on the current analog electrical signal;

[0015] After determining the current analog electrical signal and noise electrical signal, the corresponding frames of the current analog electrical signal and noise electrical signal in the induction time sequence signal are determined. This enables the subsequent optimal tap number analysis and noise reduction unit to perform Fourier transform operations on the current analog electrical signal and noise electrical signal of the corresponding frame in a precise and targeted manner, thereby obtaining the information of the two in the frequency domain more efficiently and accurately, and providing a precise basis for subsequent frequency domain-based noise reduction and other processing.

[0016] Based on the above technical solution, the present invention can be further improved as follows: the optimal tap number analysis and noise reduction unit randomly sets different numbers of taps, and uses different numbers of taps and corresponding tap coefficients to form different numbers of taps as candidate sets and corresponding tap coefficients; each tap in the candidate set is selected in turn, and the tap coefficient corresponding to the selected tap is randomly assigned an initial value of > through a random function; and the similarity between the delayed analog electrical signal and the current analog electrical signal in the same time window in different historical time windows is calculated, and the delayed analog electrical signal in the time window with the highest similarity is selected;

[0017] The program retrieves a delayed analog signal within the time window that has one less tap than the number of taps. It assigns corresponding tap coefficients to the delayed analog signal and the current analog signal. The program then multiplies the current analog signal and the delayed analog signal by their respective tap coefficients, sums all the products, and outputs the filtered signal.

[0018] As a further improvement to this technical solution, the optimal tap number analysis and noise reduction unit receives the noise electrical signal in the induced timing signal, performs a Fourier transform on the noise electrical signal to obtain the noise spectrum of the noise electrical signal, and takes the average of the squares of the noise spectrum amplitude to obtain the noise power spectrum; performs another Fourier transform on the current analog electrical signal to obtain the current spectrum; divides the value of the current spectrum by the square root of the noise power spectrum to obtain the estimated spectrum, and performs an inverse transform on the estimated spectrum to obtain the estimated noise electrical signal.

[0019] As a further improvement to this technical solution, the Fourier transform in the optimal tap number analysis and noise reduction unit is used to decompose the time-domain signal into a weighted sum of complex exponential signals of different frequencies, thereby converting the time-domain signal from the time domain to the frequency domain, revealing the frequency components contained in the time-domain signal and their corresponding amplitude and phase information; the amplitude of each frequency component is the contribution intensity of the corresponding frequency in the time-domain signal, and the phase is the starting position of the corresponding frequency; the frequency domain is equal to the time-domain signal multiplied by the negative j of the natural exponent multiplied by π multiplied by the frequency multiplied by the time power, and then integrated over time from negative infinity to positive infinity, thereby completing the conversion; the time-domain signal specifically consists of a noise electrical signal and the current analog electrical signal.

[0020] The beneficial effect of the above-mentioned further scheme is that by decomposing the time-domain signal into a weighted sum of complex exponential signals of different frequencies, the conversion from the time domain to the frequency domain can be accurately realized, clearly revealing the frequency components contained in the time-domain signal and the amplitude and phase information corresponding to each component. The amplitude reflects the contribution intensity of the corresponding frequency in the time-domain signal, and the phase reflects the starting position of the corresponding frequency. Furthermore, the specific calculation method for the conversion from the time domain to the frequency domain is clarified, and the time-domain signal for this transformation is determined to be a noise electrical signal and the current analog electrical signal.

[0021] Based on the above technical solution, the present invention can be further improved as follows: the optimal tap number analysis and noise reduction unit is composed of the current analog electrical signal and the estimated noise electrical signal superimposed on the induction timing signal. At this time, the expected signal is obtained by subtracting the estimated noise electrical signal from the current analog electrical signal, and the error signal is obtained by subtracting the filtered signal from the expected signal.

[0022] Based on minimizing the error signal, multiple error signals are obtained by multiplying the error signal by the step size factor and then by the delayed analog electrical signal, with different randomly set tap coefficients. The error signal with the smallest value is selected as the error signal for the corresponding number of taps. The number of taps in the candidate set is selected again, and the error signal is calculated repeatedly. The error signals corresponding to each number of taps are compared, and the optimal number of taps that minimizes the error signal and the corresponding tap coefficient are selected as the final result.

[0023] The beneficial effect of the above-mentioned further scheme is that by performing calculations on different randomly set tap coefficients in combination with step size factors, error signals, and delayed analog electrical signals, multiple error signals are obtained. The minimum error signal corresponding to the number of taps is selected from these signals. Then, candidate tap numbers are selected again and the error signals are calculated repeatedly. The error signals corresponding to each tap number are compared, and finally, the optimal number of taps and corresponding tap coefficients that minimize the error signals are selected. This accurately determines the optimal number of taps and tap coefficients, and can more effectively process the inductive time-series signal formed by the superposition of the current analog electrical signal and the estimated noise electrical signal.

[0024] Based on the above technical solution, the present invention can be further improved as follows: the signal amplification and optimization module receives the optimal number of taps corresponding to the current analog electrical signal, multiplies the optimal number of taps by the basic proportional coefficient to obtain the amplification coefficient; then it receives the filtered signal after noise reduction and filtering, and multiplies the amplification factor by the filtered signal to obtain the amplified signal.

[0025] The beneficial effect of the above-mentioned further scheme is that by prioritizing the determination of the optimal number of taps adapted to the current analog electrical signal, and using this as a basis to generate a suitable amplification factor in conjunction with the basic proportional coefficient, the amplitude of the noise-reduced and filtered signal can be amplified. This allows for precise determination of the amplification factor based on the characteristics of the current analog electrical signal, and targeted amplification of the filtered signal. This ensures that the signal quality is improved after noise reduction, and that the signal amplitude meets the requirements of subsequent transmission and use through reasonable amplification, effectively enhancing the usability and effectiveness of the signal. This, in turn, better guarantees the signal processing and communication effects of radio communication systems and other related equipment with intercom functions.

[0026] A radio communication method with intercom functionality includes the following steps:

[0027] S1. Collect ambient voice and convert it into an induced electrical signal. Define the induced electrical signal with the same sampling interval as the induced time sequence signal. Determine the noise electrical signal and the current analog electrical signal in the induced time sequence signal.

[0028] S2. Set multiple tap numbers and corresponding tap coefficients, and sequentially use different tap numbers to denoise the current analog electrical signal output filtered signal; obtain the desired signal by performing Fourier transform on the noise electrical signal in the sensing time sequence signal and the current analog electrical signal, calculate the error signal, set constraints to minimize the error signal, iterate the tap coefficients and tap number, and determine the optimal number of taps and corresponding tap coefficients.

[0029] S3. Employ the optimal number of taps and dynamically set the amplification factor for different speech features to amplify the current analog electrical signal.

[0030] In addition to the objectives, features, and advantages described above, the present invention has other objectives, features, and advantages. The invention will now be described in further detail with reference to the figures. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the overall module of the present invention;

[0032] Figure 2 This is a schematic diagram illustrating the working principle of the core processing module for induced electrical signals of the present invention.

[0033] Figure 3 This is a schematic diagram illustrating the working principle of the optimal tap count analysis and noise reduction unit and the signal amplification optimization module in this invention.

[0034] Figure 4 This is a flowchart illustrating the working steps of the present invention.

[0035] The meanings of the labels in the diagram are as follows:

[0036] 1. Environmental voice acquisition module; 2. Core processing module for induced electrical signals; 21. Signal frame energy discrimination unit; 22. Optimal tap number analysis and noise reduction unit; 3. Signal amplification and optimization module. Detailed Implementation

[0037] The technical solutions in 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.

[0038] refer to Figures 1-3As shown, a wireless communication system with intercom function includes an environmental voice acquisition module 1, an inductive electrical signal core processing module 2, and a signal amplification and optimization module 3. The environmental voice acquisition module 1 uses a microphone to acquire environmental voice. The microphone includes a diaphragm, and a lightweight coil made of extremely fine copper wire is fixedly connected to the back of the diaphragm. The coil is suspended in a constant magnetic field formed by a permanent magnet (usually a neodymium iron boron magnet), and the axial direction of the coil is perpendicular to the direction of the magnetic field lines.

[0039] When the microphone receives the drive signal (if the user presses and holds the PTT (one-button intercom) button), the ambient voice is transmitted to the microphone in the form of sound waves (essentially mechanical waves formed by the periodic changes in the density of air molecules). The sound waves of the ambient voice generate a periodic pressure difference on the microphone diaphragm. At this time, the diaphragm generates reciprocating mechanical vibrations that are completely matched with the frequency and amplitude of the ambient voice. That is, the frequency of the ambient voice determines the vibration frequency of the diaphragm.

[0040] Because the diaphragm and coil are fixedly connected, when the diaphragm vibrates, it synchronously drives the coil to reciprocate linearly with the sound wave. As the coil continuously cuts the magnetic field lines generated by the permanent magnet, an induced electromotive force is generated at the ends of the coil according to the law of electromagnetic induction. That is, the induced electrical signal that conforms to the changing patterns of environmental speech. .

[0041] Environmental voice acquisition module 1 acquires sensing timing signals. During the process, the user's output voice signal is not continuous. Normal speech is formed by the regular vibration of the vocal organs and the control of airflow. When the user speaks, pauses will be made between words and sentences according to semantic and grammatical requirements in order to achieve clear expression.

[0042] Then, during the pause interval, theoretically there should be no valid voice signal generated by the user actively speaking. At this time, if the environmental voice acquisition module 1 still acquires the sensing timing signal... Then determine the induction timing signal at this time. Noise electrical signals caused by various interference factors in the environment For example, electromagnetic interference in the surrounding environment, electronic noise from the equipment itself, and other irrelevant background noises (such as the sound of an air conditioner running, a slight electrical hum, etc.).

[0043] Furthermore, noise signals will also be present during the normal output of voice by the user. Therefore, in order to control the timing signal of the sensing Noise reduction and filtering; therefore, the core processing module 2 for induced electrical signals includes a signal frame energy discrimination unit 21 and an optimal tap number analysis and noise reduction unit 22.

[0044] The signal frame energy discrimination unit 21 receives the induced electrical signals collected at t times within the same sampling interval and defines them as induced timing signals. The sensing timing signal is divided into frames according to the frame length N and the frame shift R. After frame division, the first The frame's signal sensing timing signal is ,in , Frame index, used to identify different frames;

[0045] The first The sum of the squares of the amplitudes of all sampling points within the frame is used to obtain the first... Short-time energy of a frame ,in For the first The square of the amplitude corresponding to the nth sampling point within the frame;

[0046] Train the threshold prediction model, and then use the first... Short-time energy of a frame Input into the threshold prediction model and set the energy threshold. Then compare the energy threshold. With short-term energy If the energy threshold Short-term energy Then determine the first Intra-frame sensing timing signal Noise electrical signal If the energy threshold ≥ Short-time energy Then determine the first Intra-frame sensing timing signal For the current analog electrical signal Then, through optimal tap count analysis and noise reduction filtering by noise reduction unit 22, the current analog electrical signal is processed. .

[0047] The threshold prediction model consists of an input layer, convolutional layers, pooling layers, fully connected layers, and an output layer.

[0048] The input layer is used to normalize the short-time energy of each short-time energy (mapping the short-time energy to the interval [0,1] to avoid the impact of differences in data magnitude on training). (in , The minimum and maximum values ​​among all short-time energies are respectively used as input features and output to the convolutional layer;

[0049] The convolutional layer contains multiple convolutional kernels. These kernels slide across the input features to calculate a weighted sum, extracting local features of the short-time energy over time (such as energy abrupt changes, stationary segments, etc.). Specifically, the sequence of input features corresponding to the normalized short-time energy of the received input is... ( (where the sequence length is 1), multiple convolutional kernels are... ( (where is the kernel length), and the stride of the kernel sliding across the input features is... For the first Each convolutional kernel performs a convolution operation on the input sequence and outputs a feature map. The element ,in This represents the normalized short-time energy value at the corresponding position in the input sequence. For the first The nth convolutional kernel The weight value of each position;

[0050] Max pooling downsamples the local feature maps output by the convolutional layers, reducing dimensionality while preserving key features. ,in Step size, This is the pooling window size;

[0051] Fully connected layers integrate all local feature maps to form a comprehensive feature vector: ,in This is the weight matrix of the fully connected layer. For bias vectors, This is the output of the pooling layer;

[0052] The output layer uses the softmax activation function to output short-time energy, which is a noisy electrical signal. Current analog electrical signal Corresponding output probability: ,in For the first The predicted probabilities of each category, The first output of the fully connected layer The original values ​​of each node (output without activation function processing). The first output of the fully connected layer The original values ​​of each node Perform with constant Exponential operations with base 0;

[0053] Set an initial energy threshold, compare the output probability with the initial energy threshold, and determine the prediction label to which the short-term energy belongs.

[0054] When the signal frame energy discrimination unit 21 trains the threshold prediction model, it first uses the environmental voice acquisition module 1 to acquire noise electrical signals and analog electrical signals known to be in a noisy environment. Specifically, the noise electrical signals are signal frames acquired when there is no effective voice input (such as signals containing only environmental interference, equipment noise, etc.), and the analog electrical signals are signal frames containing effective voice components (such as signals acquired when the user speaks). When the environmental voice acquisition module 1 acquires noise electrical signals and analog electrical signals known to be in a noisy environment, it can acquire them by setting the acquisition environment by the user. For example, in a quiet room where no one is speaking, the environment may contain a faint humming sound from the air conditioner, a slight current sound from the circuit in the room, or a low-pitched noise from a distant vehicle outside the window. When acquiring analog electrical signals, the user can output a voice signal (such as different voice signals like "hello") to the environmental voice acquisition module 1.

[0055] Then, the known noisy electrical signal and the analog electrical signal are divided into frames, and the short-time energy of each frame is calculated as training samples.

[0056] The training samples are divided into a training set and a validation set. During training, the training set samples are first input into the threshold prediction model for forward propagation to obtain the predicted labels.

[0057] The error between the predicted label and the true label is calculated using the cross-entropy loss function: ( (Assuming the true labels), the backpropagation algorithm is then used to propagate the error from the output layer back to each layer, calculate the gradient of each parameter (convolutional kernel weights, fully connected layer weights, initial energy threshold, etc.), and use an optimization algorithm (such as Adam) to update the parameters based on the gradients to minimize the loss. ,in For the current parameter, For learning rate, For gradient;

[0058] Through multiple rounds of iterative training (each round using different batches of samples), the initial energy threshold is made capable of accurately determining the label corresponding to short-term energy. The accuracy of the initial energy threshold is monitored using a validation set until the accuracy reaches its maximum, at which point the initial energy threshold training is considered complete. The initial energy threshold is defined as the energy threshold in the signal frame energy discrimination unit 21. .

[0059] The cross-entropy loss function is used to measure the difference between the predicted and actual results. After forward propagation calculates the prediction results of the threshold prediction model, the loss value is calculated based on the cross-entropy loss function. Using the backpropagation algorithm, the loss value is propagated back to each layer of the convolutional neural network, calculating the gradient of each parameter with respect to the loss value. The parameters in the network (such as the weights of convolutional kernels and the connection weights of fully connected layers) are updated using gradient descent (or its variants, such as stochastic gradient descent and adaptive gradient descent), gradually reducing the loss value. After multiple rounds of training (i.e., multiple forward and backpropagation cycles to update parameters), the model continuously learns the mapping relationship between the input short-time energy data and the signal category, improving the accuracy of the prediction. The threshold prediction model is constructed using the parameters of the layers with the highest prediction results, ultimately resulting in a model that can accurately set the energy threshold. Threshold prediction model;

[0060] The optimal tap number analysis and noise reduction unit 22 receives the current analog electrical signal in the signal frame energy discrimination unit 21. Randomly set different numbers of taps Q, and use different numbers of taps To form a candidate set and the corresponding tap coefficient ( ); select candidate sets in sequence Number of taps in And the selected number of taps is determined by a random function. Corresponding tap coefficient Randomly assign an initial value greater than 0; calculate the delayed analog electrical signal within different historical time windows. With current analog electrical signal Based on similarity within the same time window, the delayed analog electrical signal within the time window with the highest similarity is selected. Select multiple delayed analog electrical signals ; for the current analog electrical signal Delayed analog electrical signals Assign corresponding tap coefficients Weighted summation of the current analog electrical signal Delayed analog electrical signals Output filtered signal ,in For the first Time, Number tap coefficient , Tap coefficient The index;

[0061] Optimal tap count analysis and noise reduction unit 22 for Time before The time-delayed electrical signal reflects the current analog electrical signal. The timing characteristics, among which The range of values ​​is For example: if a tap coefficient is set If the number of taps Q=8, then The possible values ​​are 0, 1, 2, ..., 7;

[0062] In order to select the optimal number of taps from the candidate set Q and the optimal number of taps Corresponding to each tap coefficient This enables the extraction of electrical signals containing noise. The current analog electrical signal The noise is separated to improve the quality of radio communication and ensure that the induced electrical signal can be clearly received and identified. The optimal tap number analysis and noise reduction unit 22 is also used to receive the induced timing signal. Noise electrical signals in Fourier transform noise electrical signal Obtaining the noise electrical signal from the frequency domain signal In the Frame, Frequency Index Noise spectrum at (in For the first Discrete-time sequence of frame noise electrical signal For the length of each frame, For frequency domain indexing, ), for the first Frame noise spectrum The noise power spectrum is obtained by averaging the squares of the amplitudes. ;

[0063] Fourier transform the current analog electrical signal again Given a frequency domain signal, obtain the current spectrum. ,in For the first The current discrete-time sequence of the analog electrical signal;

[0064] Calculate the first Frame, Frequency Index Estimated spectrum at [location] Inverse conversion estimation spectrum The estimated noise electrical signal is obtained. ;

[0065] Optimal tap count analysis and Fourier transform noise electrical signal in noise reduction unit 22 Current analog electrical signal At this time, since the signal frame energy discrimination unit 21 has already divided the induction timing signal before this, For the current analog electrical signal and estimating noise electrical signals And determine its corresponding frame length, so that the optimal tap number analysis and noise reduction unit 22 uses Fourier transform to analyze the current analog electrical signal. and noise electrical signals During the conversion process, the short-time stationary characteristics adapted to the known frame length can be used to more accurately capture the current analog electrical signal. and noise electrical signals The frequency components within a known frame length provide more targeted and reliable spectral information support for subsequent frequency domain-based noise estimation and signal denoising operations.

[0066] Based on induction timing signal From the current analog electrical signal and estimating noise electrical signals This is achieved by superposition, and at this time, it is transmitted through the current analog electrical signal. Subtract the estimated noise signal To obtain the desired signal (excluding noise electrical signals) (pure electrical signal), through the desired signal Subtract the filtered signal Obtain the error signal ;

[0067] Based on minimizing the error signal Iteration tap coefficients ,in Step size factor controls the iteration tap coefficient Speed ​​and stability; and select the number of taps from the candidate set Q again. Repeated calculation of error signal Compare the error signals corresponding to each tap number. Select the signal that minimizes the error Optimal number of taps and the corresponding tap coefficient As a final result, the filtered current analog electrical signal Preserve the voice signal to the greatest extent possible and suppress the current analog electrical signal. The noise in the filter improves the clarity and intelligibility of voice signals in radio communications, thereby outputting a filtered signal. ;

[0068] The present invention further converts the voice signal into a current analog electrical signal using a microphone. At that time, due to the working principle of microphones (such as dynamic microphones relying on electromagnetic induction, with a weak initial induced electromotive force) and structural limitations, the current analog electrical signal... The amplitude is extremely small, typically only in the millivolt or even microvolt range, while the filtered signal... Based on current analog electrical signals Generate, therefore when a weak filtered signal During subsequent transmission, the signal is easily overwhelmed by noise due to significant attenuation caused by line resistance and electromagnetic interference. Therefore, the signal amplification and optimization module 3 receives the filtered signal after noise reduction and filtering. Using magnification factor Amplified and filtered signal To amplify the signal ;

[0069] Furthermore, it is considered that if the optimal tap count analysis and noise reduction unit 22 sets the optimal tap count... At that time, specifically based on the error signal Set the optimal number of taps And error signal Specifically, the desired signal With filtered signal The difference between them, therefore, the optimal number of taps to participate in the weighted summation. The more times there are, the more delayed analog electrical signals are present. Generate filter signal together It is better suited to steady-state speech features (such as sustained pronunciation of long vowels and slow rise and fall of intonation) - the characteristics of steady-state speech are "long duration and slow change", which requires more "historical signals" to cover its long period and "fine weighting" to fit the slow change pattern.

[0070] If the optimal number of taps is used in the weighted summation The fewer the values, the more likely there is a small amount of delayed analog electrical signal. Generate filter signal together It is better suited to transient speech features (such as instantaneous plosives of voiceless consonants and rapid pitch changes) – transient speech is characterized by “short duration and drastic changes”. Too many “historical signals” can interfere with the capture of “instantaneous changes”. Fewer taps can more accurately adapt to short-term abrupt changes.

[0071] Signal amplification and optimization module 3 for filtered signals When amplifying, because the amplitude changes of steady-state speech features are gradual (e.g., the prolongation of long vowels is "gradual"), a "continuous and smooth" amplification is required—if the amplification factor is... Insufficient amplification will result in weak and disjointed speech; if the amplification factor is insufficient... Excessive changes, on the other hand, can disrupt the natural feel of "gradual change."

[0072] Transient speech features, characterized by instantaneous bursts of amplitude (such as the plosive of the voiceless consonant "p," which is an "instantaneous peak"), require "precise and non-trailing" amplification—if the amplification factor is... Too much amplification will make the "instantaneous explosion" harsh and may even amplify the noise; if the amplification factor is too high... If it's insufficient, it will lose the "sharp, explosive feeling";

[0073] Therefore, the amplification factor in signal amplification and optimization module 3 ,in It is the basic scaling factor, used to adjust the overall range of magnification, and is set according to the actual hardware characteristics; For the current analog electrical signal in the optimal tap number analysis and noise reduction unit 22 Corresponding to the optimal number of taps;

[0074] To achieve a larger optimal tap number (Steady-state characteristics) → High amplification factor ensures the coherence and energy of steady-state speech; smaller optimal tap number (Transient features) → Small amplification factor to ensure the sharpness of transient speech and noise suppression.

[0075] Signal amplification and optimization module 3 uses amplification factor Amplified and filtered signal To amplify the signal Then, the signal is modulated and amplified again. The generated radio frequency hybrid signal, specifically the amplified signal established with various countries. The mapping relationship between the high-frequency carrier amplitude and the amplitude of the amplified signal ensures that when the amplitude of the amplified signal changes, the amplitude of the high-frequency carrier changes synchronously according to the mapping relationship. The resulting radio frequency (RF) hybrid signal is then transmitted through an antenna to realize the current analog electrical signal. Transmission in space;

[0076] At the receiving end, the radio frequency mixed signal is captured by receiving equipment (such as a receiver, antenna, etc.). Based on the previously established mapping relationship between the amplified signal and the high-frequency carrier amplitude, the current analog electrical signal is extracted from the radio frequency mixed signal. (That is, signals containing information such as voice) are played.

[0077] refer to Figure 4 As shown, a radio communication method with intercom function includes the following steps:

[0078] S1. Collect ambient voice and convert it into an induced electrical signal. Define the induced electrical signal with the same sampling interval as the induced time sequence signal. Determine the noise electrical signal and the current analog electrical signal in the induced time sequence signal.

[0079] S2. Set multiple tap numbers and corresponding tap coefficients, and sequentially use different tap numbers to denoise the current analog electrical signal output filtered signal; obtain the desired signal by performing Fourier transform on the noise electrical signal in the sensing time sequence signal and the current analog electrical signal, calculate the error signal, set constraints to minimize the error signal, iterate the tap coefficients and tap number, and determine the optimal number of taps and corresponding tap coefficients.

[0080] S3. Employ the optimal number of taps and dynamically set the amplification factor for different speech features to amplify the current analog electrical signal.

[0081] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A radio communication system with intercom function, characterized in that, include: The environmental voice acquisition module (1) uses a microphone to acquire environmental voice and converts it into an inductive electrical signal; The core processing module (2) for induced electrical signals receives induced electrical signals collected at t times within the same sampling interval and defines them as induced time-series signals; trains a threshold prediction model to judge the noise electrical signals and the current analog electrical signals in the induced time-series signals; sets multiple tap numbers and corresponding tap coefficients, and sequentially uses different tap numbers to denoise the current analog electrical signals and output filtered signals; obtains the desired signal by Fourier transforming the noise electrical signals and the current analog electrical signals in the induced time-series signals, calculates the error signal, sets constraints to minimize the error signal, iterates the tap coefficients and tap numbers, determines the optimal number of taps and corresponding tap coefficients for denoising and filtering the current analog electrical signals, and obtains the filtered signal; The signal amplification and optimization module (3) receives the optimal number of taps, dynamically sets different amplification factors for speech features based on the optimal number of taps, amplifies the filtered signal, modulates the amplified filtered signal, generates a radio frequency mixed signal, and transmits the radio frequency mixed signal into space using an antenna. The core processing module (2) for the induced electrical signal includes: The signal frame energy discrimination unit (21) receives the induced timing signal collected at t times in the same sampling interval, divides the induced timing signal into frames according to the frame length N and frame shift R, calculates the short-time energy of each frame, and determines whether the induced timing signal of each frame is the current analog electrical signal. If it is not the current analog electrical signal, it is judged as a noise electrical signal. The optimal tap number analysis and noise reduction unit (22) uses different numbers of taps and corresponding tap coefficients to filter the current analog electrical signal and output the filtered signal. It also extracts multiple delayed analog electrical signals corresponding to different tap numbers, assigns tap coefficients to the delayed analog electrical signals, and outputs the filtered signal by weighted summation. It calculates and estimates the noise electrical signal based on the current analog electrical signal and the noise electrical signal, and calculates the desired signal. It iterates the tap coefficients and tap number based on the desired signal to select the optimal number of taps and tap coefficients. The signal frame energy discrimination unit (21) sums the squares of the amplitudes of all sampling points of the induced timing signal in each frame after framing to obtain the short-time energy of the induced timing signal in each frame. The threshold prediction model is trained, and the short-time energy of each frame is input into the threshold prediction model to set the energy threshold. Then the energy threshold and the short-time energy are compared: if the energy threshold is less than the short-time energy, the induction timing signal in the frame is judged to be a noise electrical signal; otherwise, the induction timing signal in the frame is judged to be the current analog electrical signal. Then the current analog electrical signal is denoised and filtered by the optimal tap number analysis and the denoising unit (22). When training the threshold prediction model, the signal frame energy discrimination unit (210) first uses the environmental speech acquisition module (1) to acquire known noise electrical signals and analog electrical signals in a noisy environment; by dividing the known noise electrical signals and analog electrical signals into frames, the short-time energy of each frame is calculated as a training sample; The training samples are divided into a training set and a validation set. During training: the training set samples are forward propagated to obtain the predicted label; the error between the predicted label and the real label is calculated; the error is propagated back from the output layer to each layer, and the gradient of each parameter is calculated to minimize the loss and update the parameters; the accuracy of the initial energy threshold is monitored through the validation set until the accuracy is maximized, then the initial energy threshold training is considered complete. The initial energy threshold is defined as the energy threshold in the signal frame energy discrimination unit (21). The optimal tap number analysis and noise reduction unit (22) receives the current analog electrical signal output by the signal frame energy discrimination unit (21); A candidate set and corresponding tap coefficients are randomly set; each tap number in the candidate set is selected in turn, and the tap coefficients corresponding to the selected tap numbers are randomly assigned an initial value >0 by a random function; the similarity between the delayed analog electrical signal and the current analog electrical signal in the same time window in different historical time windows is calculated, and the delayed analog electrical signal in the time window with the highest similarity is selected. Retrieve the delayed analog electrical signal within the time window, assign corresponding tap coefficients to the delayed analog electrical signal and the current analog electrical signal, multiply the current analog electrical signal and the delayed analog electrical signal by the corresponding tap coefficients respectively, and add all the products to obtain the filtered signal and output it.

2. The radio communication system with intercom function according to claim 1, characterized in that: The microphone in the environmental voice acquisition module (1) includes a diaphragm, and a coil is fixed on the back of the diaphragm; the coil is suspended in a constant magnetic field formed by a permanent magnet, and the axial direction of the coil is perpendicular to the magnetic field lines. After the microphone receives the drive signal, the ambient voice propagates to the microphone in the form of sound waves. The sound waves of the ambient voice act on the diaphragm, and the diaphragm generates a periodic pressure difference and reciprocating mechanical vibration. When the diaphragm vibrates, it synchronously drives the coil to reciprocate linearly with the sound waves. The coil continuously cuts the magnetic field lines generated by the permanent magnet. According to the law of electromagnetic induction, an induced electromotive force is generated at both ends of the coil, which is an induced electrical signal that is consistent with the changing pattern of the ambient voice.

3. The radio communication system with intercom function according to claim 1, characterized in that: The optimal tap number analysis and noise reduction unit (22) receives the noise electrical signal in the induced timing signal, Fourier transforms the noise electrical signal into a frequency domain signal to obtain the noise spectrum of the noise electrical signal, and averages the square of the amplitude of the noise spectrum to obtain the noise power spectrum. The current analog electrical signal is transformed into a frequency domain signal by Fourier transform again to obtain the current spectrum; the estimated spectrum is obtained by dividing the current spectrum value by the square root of the noise power spectrum, and the estimated noise electrical signal is obtained by inverse transformation of the estimated spectrum.

4. The radio communication system with intercom function according to claim 3, characterized in that: The Fourier transform in the optimal tap number analysis and noise reduction unit (22) is used to decompose the time-domain signal into a weighted sum of complex exponential signals of different frequencies, and to convert the time-domain signal to the frequency domain: the amplitude of each frequency component is the contribution intensity of the corresponding frequency in the time-domain signal, and the phase is the starting position of the corresponding frequency; the frequency domain is equal to the time-domain signal multiplied by the negative j of the natural exponent multiplied by 2π multiplied by the frequency multiplied by the time power, and then integrated over time from negative infinity to positive infinity, thereby completing the conversion; the time-domain signal is specifically a noise electrical signal and the current analog electrical signal.

5. The radio communication system with intercom function according to claim 4, characterized in that: The optimal tap number analysis and noise reduction unit (22) is based on the induction timing signal being superimposed from the current analog electrical signal and the estimated noise electrical signal. At this time, the expected signal is obtained by subtracting the estimated noise electrical signal from the current analog electrical signal, and the error signal is obtained by subtracting the filtered signal from the expected signal. Based on minimizing the error signal, multiple error signals are obtained by multiplying the error signal by the step size factor and then by the delayed analog electrical signal, with different randomly set tap coefficients. The error signal with the smallest value is selected as the error signal for the corresponding number of taps. The number of taps in the candidate set is selected again, and the error signal is calculated repeatedly. The error signals corresponding to each number of taps are compared, and the optimal number of taps that minimizes the error signal and the corresponding tap coefficient are selected as the final result.

6. The radio communication system with intercom function according to claim 5, characterized in that: The signal amplification and optimization module (3) receives the optimal number of taps corresponding to the current analog electrical signal, multiplies the optimal number of taps by the basic proportional coefficient to obtain the amplification coefficient, and then receives the filtered signal after noise reduction and filtering, and multiplies the amplification factor by the filtered signal to obtain the amplified signal.

7. A radio communication method with intercom function, applied to the radio communication system with intercom function as described in any one of claims 1-6, characterized in that, Includes the following steps: S1. Collect ambient voice and convert it into an induced electrical signal. Define the induced electrical signal with the same sampling interval as the induced time sequence signal. Determine the noise electrical signal and the current analog electrical signal in the induced time sequence signal. S2. Set multiple tap numbers and corresponding tap coefficients, and sequentially use different tap numbers to denoise the current analog electrical signal output filtered signal; obtain the desired signal by performing Fourier transform on the noise electrical signal in the sensing time sequence signal and the current analog electrical signal, calculate the error signal, set constraints to minimize the error signal, iterate the tap coefficients and tap number, and determine the optimal number of taps and corresponding tap coefficients. S3. Employ the optimal number of taps and dynamically set the amplification factor for different speech features to amplify the current analog electrical signal.

Images