Self-adaptive anti-interference adjustment method and system for Bluetooth headset

Abstract

The invention discloses a self-adaptive anti-interference adjustment method and system for a Bluetooth headset, and the method comprises the steps: carrying out the normalization processing of each extracted feature, calculating a single-dimensional interference coefficient based on the normalized features, obtaining a comprehensive interference grade coefficient, and carrying out the comparison of the comprehensive interference grade coefficient and a preset threshold value, dividing interference levels; forming a state space by using the comprehensive interference level coefficient, the current signal-to-noise ratio, the current packet loss rate and the current data transmission delay, forming an action space by using adjustable parameters of the Bluetooth headset, and constructing a reward function by using the adjusted performance improvement effect to determine an optimal adjustment action set; issuing adjustment parameters corresponding to the optimal adjustment action set to a corresponding function module of the Bluetooth headset, continuously monitoring a preset period number, and then taking an average value as the adjusted performance; and calculating a reward value based on the difference between the performance after adjustment and the performance before adjustment, updating a state and action value table according to the reward value, realizing strategy iteration, and forming closed-loop feedback optimization.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication anti-interference technology, and relates to a Bluetooth headset adaptive anti-interference adjustment method and system. Background Technology

[0002] With the popularization of wireless communication technology, Bluetooth headsets, as portable audio devices, are widely used in daily commutes, sports, and office scenarios. However, Bluetooth headsets operate in the 2.4GHz ISM public frequency band, which is susceptible to electromagnetic interference from Wi-Fi signals, other Bluetooth devices, and household appliances such as microwave ovens. This can lead to audio stuttering, increased packet loss rate, and increased data transmission latency, seriously affecting the user experience.

[0003] Furthermore, existing technologies suffer from the following core defects: Limited application scenarios: analysis is only performed on preset test scenarios, failing to cover complex scenarios involving dynamic changes in interference source type, intensity, and frequency during actual use; Lack of real-time adjustment capability: outputs only performance evaluation results or grading conclusions, without providing real-time adjustment solutions for headphone hardware / software parameters, thus failing to resolve the contradiction of "passing tests but experiencing poor actual performance"; Coarse interference level classification: existing technologies mostly adopt a two-level intensity classification of "Level 1 / Level 2," which cannot accurately match the adjustment needs under different interference scenarios, easily leading to over- or under-adjustment; Fixed strategies: lacking self-learning capabilities, adjustment strategies rely on preset rules, failing to optimize adjustment effects based on interference patterns observed over long-term use. Summary of the Invention

[0004] To address the problems existing in the background technology, this invention proposes an adaptive anti-interference adjustment method and system for Bluetooth headsets, aiming to improve the audio quality, connection stability, and data transmission reliability of Bluetooth headsets in actual use.

[0005] The first aspect of this application provides a Bluetooth headset adaptive anti-interference adjustment method, including:

[0006] Intensity features, frequency features, and complexity features are extracted from the preprocessed signal, and each extracted feature is normalized to obtain normalized features.

[0007] Based on the normalized features, a single-dimensional interference coefficient is calculated to obtain a comprehensive interference level coefficient. The interference level is then classified by comparing the comprehensive interference level coefficient with a preset threshold. The interference level is used to provide the decision-making model with interference level classification and strategy optimization of the state and action value table.

[0008] The state space is constructed using the comprehensive interference level coefficient, the current signal-to-noise ratio, the current packet loss rate, and the current data transmission delay. The action space is constructed using the adjustable parameters of the Bluetooth headset. The reward function is constructed using the performance improvement effect after adjustment to determine the optimal set of adjustment actions.

[0009] The adjustment parameters corresponding to the optimal set of adjustment actions are sent to the corresponding functional module of the Bluetooth headset for continuous monitoring for a preset number of cycles, and the average value is taken as the performance after adjustment.

[0010] The reward value is calculated based on the difference between the adjusted performance and the performance before adjustment. The state and action value table is updated according to the reward value to realize strategy iteration and form a closed-loop feedback optimization.

[0011] Optionally, intensity features, frequency features, and complexity features are extracted from the preprocessed signal, including:

[0012] The Bluetooth headset uses its built-in radio frequency receiver to collect interference signals from the surrounding environment in real time at a preset collection period to obtain the original signal sequence.

[0013] The original signal sequence is subjected to filtering and denoising preprocessing to obtain the preprocessed signal;

[0014] The preprocessed signal is subjected to a Fast Fourier Transform to calculate the main interference frequency and frequency fluctuation range to extract intensity features, frequency features, and complexity features. The intensity features include peak intensity and average intensity, the frequency features include the main interference frequency and frequency fluctuation range, and the complexity features include fractal dimension and Hearst exponent. The main interference frequency is the frequency corresponding to the maximum value of the frequency domain signal amplitude spectrum, and the frequency fluctuation range is the difference between the maximum and minimum frequencies of the frequency domain signal.

[0015] Optionally, calculating the single-dimensional interference coefficient based on the normalized features includes:

[0016] The formula for calculating the intensity anomaly coefficient is:

[0017] ;

[0018] in, ; The The intensity anomaly coefficient, This is the sum of the energies of all high-frequency components after performing a discrete wavelet transform on the preprocessed signal. This is the sum of the total energy of the high-frequency components and the energy of the low-frequency components of the preprocessed signal after discrete wavelet transform. The sum of the energies of all low-frequency components after discrete wavelet transform of the preprocessed signal is given by J, where J is the preset total number of wavelet decomposition levels, and j is any wavelet decomposition level. The energy of the high-frequency component of the j-th layer;

[0019] The formula for calculating the frequency anomaly coefficient is:

[0020] ;

[0021] in, ; For frequency anomaly coefficients, The standard deviation of the frequency component amplitude, Let M be the average amplitude of the frequency components, M be the number of frequency points, and k be any frequency point. This represents the frequency domain signal value at the k-th frequency point;

[0022] The duration coefficient is calculated as follows:

[0023] ;

[0024] in, Here, T is the duration coefficient, where T is the duration of the current interference. This is the short-term interference threshold. This is the long-term interference threshold;

[0025] The formula for calculating the complexity coefficient is:

[0026] ;

[0027] in, is the complexity coefficient, D is the fractal dimension, and H is the Hearst exponent.

[0028] Optionally, the formula for calculating the comprehensive interference level coefficient is as follows:

[0029] ;

[0030] in, The comprehensive interference level coefficient is denoted by 'a', where 'a' represents the weight of the intensity anomaly coefficient. Let be the intensity anomaly coefficient, and b be the weight of the frequency anomaly coefficient. Here, c represents the frequency anomaly coefficient, and c represents the weight of the duration coefficient. The duration coefficient and the weight of the complexity coefficient d are the following: is the complexity coefficient.

[0031] Optionally, the process involves constructing a state space using the comprehensive interference level coefficient, the current signal-to-noise ratio, the current packet loss rate, and the current data transmission delay; constructing an action space using the adjustable parameters of the Bluetooth headset; and building a reward function based on the performance improvement effect after adjustment to determine the optimal set of adjustment actions, including:

[0032] The process of defining a spatial state is as follows:

[0033] ;

[0034] Where S represents the spatial state, SNR represents the current signal-to-noise ratio, and L represents the packet loss rate. Current data transmission delay;

[0035] The process of defining the reward function is as follows:

[0036] ;

[0037] in, ; ; , The difference between the adjusted signal-to-noise ratio and the original signal-to-noise ratio. The difference between the adjusted packet loss rate and the original packet loss rate is used to calculate the packet loss rate. This is the difference between the adjusted data transmission delay and the original data transmission delay.

[0038] The optimization strategy process is as follows:

[0039] ;

[0040] in, Let S be the value of each action in different states corresponding to the current state S and action A. To update the value of each action in different states corresponding to action A before the update, For learning rate, As a discount factor, The next state after adjustment, This is an optional action for the next state. It represents the maximum value among all possible actions in the next state, corresponding to the values ​​of each action in different states.

[0041] Optionally, the specific process of determining the optimal set of adjustment actions by analyzing the game relationship between the headphones and the interference source using Nash equilibrium theory includes: treating the headphones and the interference source as two parties in a game, with the headphones aiming to maximize performance and the interference source aiming to minimize headphone performance; calculating the payoff matrix under different actions, where each element in the payoff matrix represents the headphone's performance value under the corresponding action; finding the Nash equilibrium point in the payoff matrix, where the Nash equilibrium point represents an action that the headphones cannot further improve by unilaterally changing their actions; and selecting the action corresponding to the Nash equilibrium point as the preferred action and incorporating it into the optimal set of adjustment actions.

[0042] Optionally, the specific process of updating the state and action value table based on the reward value to achieve policy iteration includes: when the reward value is greater than or equal to 0.5, increasing the value weight of each action in different states corresponding to the current state and action; when the reward value is greater than 0 and less than 0.5, normally updating the value of each action in different states corresponding to the current state and action; when the reward value is less than or equal to 0, decreasing the value weight of each action in different states corresponding to the current state and action, and triggering action mutation, randomly selecting untried actions; updating the state and action value table once after each adjustment parameter execution and performance monitoring is completed, thus achieving policy iteration.

[0043] Optionally, the optimal set of adjustment actions is determined by analyzing the game relationship between the headphones and the interference source using reinforcement learning combined with Nash equilibrium theory.

[0044] Optionally, the specific process of sending the adjustment parameters to the corresponding functional module of the Bluetooth headset for execution and collecting the Bluetooth headset performance indicators includes:

[0045] The adjustment parameters are sent to the RF module, decoding module, and antenna module. The RF module adjusts the RF gain and frequency hopping frequency according to the adjustment parameters. The decoding module switches the decoding algorithm mode and adjusts the error concealment depth according to the adjustment parameters. The antenna module adjusts the antenna matching degree according to the adjustment parameters.

[0046] A second aspect of this application provides a Bluetooth headset adaptive anti-interference adjustment system, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the aforementioned Bluetooth headset adaptive anti-interference adjustment method.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] This invention provides a Bluetooth headset adaptive anti-interference adjustment method and system with significant and multi-dimensional beneficial effects, as detailed below:

[0049] The multi-dimensional interference level classification adjustment mechanism enables targeted optimization based on interference of varying intensity, frequency, and complexity, allowing the headphones to achieve optimal performance under diverse interference scenarios and precisely meet users' varied needs. Employing a decision-making mechanism combining reinforcement learning and Nash equalization, the headphones are endowed with the ability to autonomously learn the optimal adjustment strategy. With prolonged use, the headphone's adjustment effect continuously optimizes, automatically adapting to different usage environments and interference conditions without manual intervention, significantly enhancing the product's intelligence and user convenience. In heavily interfered scenarios, this invention delivers multifaceted performance improvements: a significantly improved signal-to-noise ratio, enhancing signal quality and providing users with clearer, purer sound; a reduced packet loss rate, ensuring complete audio data transmission and avoiding intermittent sound issues; and reduced data transmission latency, achieving smoother audio playback and significantly improving the overall user experience.

[0050] In summary, the technical solution of this patent application performs excellently in terms of real-time performance, accuracy, intelligence, versatility, and performance improvement, and has broad application prospects and significant economic and social benefits. Attached Figure Description

[0051] Figure 1 This is a flowchart of a Bluetooth headset adaptive anti-interference adjustment method according to an embodiment of the present invention;

[0052] Figure 2 This is a schematic diagram of a Bluetooth headset adaptive anti-interference adjustment system according to an embodiment of the present invention. Detailed Implementation

[0053] 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.

[0054] In one embodiment, such as Figure 1 As shown, an adaptive anti-interference adjustment method for Bluetooth headsets is provided, which is applied to... Figure 1 Taking China as an example, the following specific steps will be used:

[0055] S10: Extract intensity features, frequency features, and complexity features from the preprocessed signal, and normalize each extracted feature to obtain normalized features.

[0056] Specifically, real-time interference signal perception and feature extraction are fundamental steps for Bluetooth headsets to achieve adaptive anti-interference adjustment. By continuously collecting interference signals in the environment and extracting key features, data support is provided for subsequent interference level classification and adjustment decisions. The specific process is as follows:

[0057] First, the Bluetooth headset's built-in radio frequency (RF) receiver unit performs real-time acquisition of interference signals. This RF receiver unit supports 2.4GHz band signal reception and continuously captures interference signals from the surrounding environment at a preset acquisition period, forming a raw signal sequence after each acquisition. This raw signal sequence refers to the set of signal data directly obtained from the environment without any processing; it includes raw information from various interference sources in the environment, such as Wi-Fi signals, signals from other Bluetooth devices, and electromagnetic signals from household appliances like microwave ovens.

[0058] Subsequently, the acquired raw signal sequence underwent filtering and denoising preprocessing. Since the raw signal sequence may contain high-frequency noise, which can interfere with the accuracy of subsequent feature extraction, a filtering algorithm is needed to process it. Specifically, a Butterworth low-pass filter algorithm is used to filter the raw signal sequence, removing high-frequency noise components by setting a cutoff frequency, ultimately obtaining the preprocessed signal. Compared to the raw signal sequence, the preprocessed signal has significantly reduced noise components and more accurately reflects the characteristics of effective interference signals in the environment.

[0059] Next, multi-dimensional feature extraction is performed on the preprocessed signal, including intensity, frequency, and complexity features. Intensity features characterize the energy level of the interference signal, specifically peak intensity and average intensity. Peak intensity refers to the maximum intensity value of the preprocessed signal within the acquisition period. Average intensity is the arithmetic mean of the intensity values ​​of all sampling points within the acquisition period, calculated as the sum of the intensity values ​​of all preprocessed signals within the acquisition period divided by the number of sampling points. Frequency features characterize the frequency distribution characteristics of the interference signal, specifically the main interference frequency and frequency fluctuation range. The main interference frequency is the frequency corresponding to the maximum amplitude spectrum value in the frequency domain signal after performing a Fast Fourier Transform on the preprocessed signal. The frequency fluctuation range is the difference between the maximum and minimum frequencies in the frequency domain signal. Complexity features characterize the structural complexity of the interference signal, specifically fractal dimension and Hearst exponent. The fractal dimension is calculated using the box counting method, with a value ranging from 1 to 2; a larger value indicates a more complex structure of the interference signal. The Hearst exponent is calculated through Hearst exponent analysis. Its value ranges from 0 to 1. When the value is greater than 0.5, it indicates that the interference signal has a long-term positive correlation, that is, the interference has a continuous and stable characteristic.

[0060] Finally, the extracted intensity, frequency, and complexity features are normalized. Since different features have different dimensions and numerical ranges, direct use would affect the accuracy of subsequent analysis. Therefore, each feature value needs to be mapped to the interval between 0 and 1 to obtain normalized features. The normalization formula is: normalized feature value equals the original feature value minus the theoretical minimum value of the feature divided by the theoretical maximum value minus the theoretical minimum value of the feature. Here, the original feature value is the extracted, unnormalized feature value, and the theoretical maximum and minimum values ​​are the maximum and minimum values ​​that the feature may have in practical application scenarios, respectively. Through normalization, the dimensional differences between different features can be eliminated, allowing various features to be analyzed and compared on the same scale.

[0061] S20: Calculate the single-dimensional interference coefficient based on the normalized features and obtain the comprehensive interference level coefficient, so as to classify the interference level by comparing the comprehensive interference level coefficient with the preset threshold; wherein, the interference level is used to provide the decision model with interference level classification and strategy optimization of state and action value table.

[0062] Specifically, interference level classification is a crucial step in determining the severity of interference by quantifying its multi-dimensional characteristics, based on real-time perception and feature extraction of interference signals. Its core is to transform normalized features into interference levels that can be directly used for adjustment decisions using a multi-dimensional representation method. The specific process is as follows:

[0063] First, a single-dimensional interference coefficient is calculated based on normalized features. This single-dimensional interference coefficient includes an intensity anomaly coefficient, a frequency anomaly coefficient, a duration coefficient, and a complexity coefficient. These four coefficients quantify the interference characteristics from four dimensions: intensity fluctuation, frequency fluctuation, duration, and structural complexity.

[0064] The calculation of the intensity anomaly coefficient requires first performing a discrete wavelet transform on the preprocessed signal, decomposing it into high-frequency and low-frequency components according to a preset number of layers (e.g., 3 layers), then calculating the sum of the energies of all high-frequency components and the total energy, and finally applying the intensity anomaly coefficient calculation formula. Obtain the intensity anomaly coefficient The intensity anomaly coefficient is used to characterize the severity of fluctuations in the intensity of interference signals.

[0065] The calculation of the frequency anomaly coefficient requires the frequency domain signal obtained by performing a Fast Fourier Transform on the preprocessed signal. First, the amplitude spectrum of the frequency domain signal is extracted, and then the standard deviation of the frequency component amplitudes is calculated. The average value of the frequency component amplitude .in, ; For frequency anomaly coefficients, The standard deviation of the frequency component amplitude, Let M be the average amplitude of the frequency components, M be the number of frequency points, and k be any frequency point. This represents the frequency domain signal value at the k-th frequency point. Then, the frequency anomaly coefficient is calculated using the formula... Obtain the frequency anomaly coefficient The frequency anomaly coefficient is used to characterize the fluctuation amplitude of the interference signal frequency.

[0066] The calculation of the duration coefficient requires first obtaining the duration T of the current interference, and then comparing T with the short-term interference threshold. , The relationship between long-term interference thresholds is determined. When Duration coefficient =0; when Duration coefficient ;when Duration coefficient =1, this coefficient is used to characterize the duration of the interference signal.

[0067] The calculation of the complexity coefficient needs to be based on the previously extracted complexity features, using the complexity coefficient calculation formula. Obtain the complexity coefficient ,in, Here, is the complexity coefficient, D is the fractal dimension, and H is the Hearst exponent. The complexity coefficient is used to characterize the structural complexity of the interference signal.

[0068] Secondly, a weighted summation method is used to integrate the single-dimensional interference coefficients to obtain the comprehensive interference level coefficient. The integration process requires assigning preset weights to the four single-dimensional interference coefficients: weight a = 0.3 for the intensity anomaly coefficient, weight b = 0.25 for the frequency anomaly coefficient, weight c = 0.2 for the duration coefficient, and weight d = 0.25 for the complexity coefficient. Then, the comprehensive interference level coefficient is calculated using the formula... The overall interference level coefficient was calculated. The comprehensive interference level coefficient is denoted by 'a', where 'a' represents the weight of the intensity anomaly coefficient. Let be the intensity anomaly coefficient, and b be the weight of the frequency anomaly coefficient. Here, c represents the frequency anomaly coefficient, and c represents the weight of the duration coefficient. The duration coefficient and the weight of the complexity coefficient d are the following: This represents the complexity coefficient. Weighted summation balances the impact of interference characteristics across different dimensions on the overall interference level, ensuring that the overall interference level coefficient comprehensively reflects the overall severity of the interference.

[0069] Finally, the interference levels are classified based on the comparison between the comprehensive interference level coefficient and the preset threshold. The preset threshold was determined through extensive experimental verification in numerous interference scenarios, and the specific classification criteria are as follows: when the comprehensive interference level coefficient C < 0.2, the interference level is classified as mild interference; when 0.2 ≤ C < 0.5, the interference level is classified as moderate interference; when 0.5 ≤ C < 0.8, the interference level is classified as severe interference; and when C ≥ 0.8, the interference level is classified as extreme interference. This classification method transforms abstract interference characteristics into clear level classifications, providing a clear basis for subsequent adaptive adjustment decisions and ensuring that the adjustment strategy accurately matches the severity of the interference.

[0070] In this invention, classifying interference levels is a crucial bridge connecting the comprehensive interference level coefficient and the adaptive adjustment decision of the Bluetooth headset. It transforms abstract, quantifiable values ​​into clear, actionable scenario identifiers, providing core support for the implementation of adjustment strategies. On one hand, classifying interference levels provides a precise scenario matching benchmark for the adjustment strategy. Different interference levels correspond to different anti-interference requirements. Mild interference only requires slight parameter adjustments to balance performance and power consumption, while extreme interference necessitates switching anti-interference modes and deeply adjusting hardware parameters. By classifying interference levels, the correlation between interference scenarios and the adjustment strategy framework can be directly established, avoiding the problem of mismatch between adjustment intensity and interference severity. On the other hand, classifying interference levels simplifies decision-making logic and reduces computational complexity. The comprehensive interference level coefficient is a continuous value within the range of 0-1. If decisions are made directly based on this value, an infinite number of numerical scenarios need to be processed. This leads to complex algorithms and response delays. However, by classifying interference levels, continuous values ​​can be discretized into four finite levels: mild, moderate, severe, and extreme. Each level corresponds to a standardized decision-making logic, which significantly reduces the amount of computation and ensures that adjustment decisions can respond to dynamic interference in real time. At the same time, classifying interference levels can also clarify the priority of performance optimization. The focus of headphone performance optimization is different under different levels. For example, under mild interference, the priority is to control power consumption, while under severe interference, the priority is to ensure connection stability. Classifying interference levels can clearly define the optimization direction for each scenario and avoid the problem of over-adjusting power for mild interference or under-adjusting for extreme interference, which can lead to performance crashes.

[0071] In this invention, the interference level classification defines the core scene dimension of the state space, providing a basic anchor point for the mapping of "state and action" in the state-action value table. The state-action value table is a mapping table of "state, action, and value," where "state" must contain complete information on "interference environment + headphone performance." The interference level classification is precisely the core quantitative classification of the "interference environment," meaning the interference level is derived from the comprehensive interference level coefficient, which is a key component parameter of the spatial state. In other words, each interference level corresponds to a set of states with similar interference characteristics; for example, "moderate interference" corresponds to all states with a comprehensive interference level coefficient of 0.2 ≤ C < 0.5. Each "state-action" value record in the state-action value table quantifies the effectiveness of a specific action under a certain interference level. Without interference level classification, the "interference environment" dimension in the state space will only be a continuous comprehensive interference level coefficient, which would require the state and action value table to maintain an infinite number of value records of similar states, making it redundant and difficult to converge. However, interference level classification, by discretizing continuous coefficients into a finite number of levels, allows the state and action value table to reuse the "action and value" logic of similar states according to the level, laying the foundation for the efficient construction and updating of the state and action value table.

[0072] The interference levels are used to provide the decision-making model with interference level classification and strategy optimization for the state and action value table. The execution results of the adjustment actions matched by the interference level classification provide key reward value basis for updating the state and action value table. The purpose of interference level classification is to match differentiated adjustment actions for different interference scenarios. For example, mild interference corresponds to "RF gain fine-tuning", and severe interference corresponds to "frequency hopping, decoding mode switching, and antenna matching optimization". The execution effect of these adjustment actions is directly quantified by the reward value, which is calculated from the difference in signal-to-noise ratio, packet loss rate, and data transmission delay before and after adjustment. The reward value is the input parameter in the state and action value table update formula. Specifically, when an adjustment action is executed under a certain interference level, if the reward value is high (≥0.5), the weight of the state and action value value corresponding to that interference level will be increased when updating the state and action value table; if the reward value is low (≤0), the weight of the corresponding state and action value will be reduced and action mutation will be triggered. It can be said that the interference level classification determines which actions to test, while the reward value after the action is executed determines how to update the value of these actions in the state and action value table. Together, they constitute the action selection and effect feedback chain for updating the state and action value table.

[0073] The iterative results of updating the state and action value table have a feedback effect on the selection of adjustment actions in subsequent scenarios with the same interference level, optimizing the matching accuracy between interference level and adjustment strategy. The essence of updating the state and action value table is to optimize the value ranking of actions under different states through historical adjustment experience. For example, if the state and action value values ​​of an action under a certain interference level continuously increase after multiple executions, it indicates that the effectiveness of the action at that level has been verified. When encountering interference scenarios of the same level again, the decision model will prioritize this action. For instance, under a moderate interference level, the action "RF gain + 2 dB + frequency hopping to 2.450 GHz" has significantly higher state and action value values ​​than other actions after multiple executions. After updating the state and action value table, when encountering moderate interference in the future, the decision model will directly prioritize this action, without needing to repeatedly explore inefficient actions. A closed loop is formed, which includes interference level classification, action selection, execution feedback, updating the state and action value table, and optimizing the action selection for that level. This makes the matching accuracy between interference level and adjustment action continuously improve with the number of iterations, and finally realizes a stable strategy for the optimal adjustment action corresponding to a specific interference level. This allows the scenario classification provided by the interference level classification and the strategy optimization record provided by the state and action value table to be recorded.

[0074] S30: The state space is composed of the comprehensive interference level coefficient, the current signal-to-noise ratio, the current packet loss rate, and the current data transmission delay. The action space is composed of the adjustable parameters of the Bluetooth headset. The reward function is constructed based on the performance improvement effect after adjustment to determine the optimal set of adjustment actions.

[0075] Specifically, adaptive adjustment decision-making is the core step in determining the optimal anti-interference adjustment scheme for Bluetooth headphones based on multi-dimensional interference level classification results. It achieves intelligent response to dynamic interference by constructing a reinforcement learning-Nash equalization fusion model. The specific process is as follows:

[0076] First, a fusion model of reinforcement learning and Nash equilibrium is constructed. The core of this model is to clarify the logical relationships between state, action, and reward, thereby supporting the learning and optimization of subsequent decision-making strategies. The state space is composed of a comprehensive interference level coefficient, current signal-to-noise ratio (SNR), current packet loss rate, and current data transmission delay. The comprehensive interference level coefficient quantifies the overall severity of interference in the current environment; the current SNR reflects the purity of the signal received by the headset; the current packet loss rate reflects the proportion of data packets lost during data transmission; and the current data transmission delay characterizes the time difference between the sender and receiver. The combination of these four parameters comprehensively describes the current state of the interference environment and headset performance, providing fundamental information for decision-making. The action space consists of adjustable parameters of the Bluetooth headset, specifically including the adjustment parameters of the RF module, decoding module, and antenna module. The adjustable parameters of the radio frequency (RF) module include RF gain and frequency hopping frequency. RF gain changes the strength of the signal received by the headphones, while frequency hopping frequency allows the headphones to switch operating frequencies to avoid interference bands. The adjustable parameters of the decoding module include decoding algorithm mode and error concealment depth. The decoding algorithm mode can switch to a more suitable decoding logic based on interference conditions, while error concealment depth improves the ability to compensate for lost data. The adjustable parameters of the antenna module include antenna matching degree, which optimizes the efficiency of antenna signal reception. These adjustable parameters together constitute the set of actions the headphones take to cope with interference. The reward function is constructed based on the performance improvement after adjustment. Its role is to quantify the effectiveness of each adjustment action, determining the reward value by comparing the performance difference before and after adjustment, ensuring that the model can distinguish between effective and ineffective actions based on the reward value.

[0077] Secondly, a state-action value learning algorithm is used to train the decision-making strategy and update the value of each action in different states. This algorithm is a value-based reinforcement learning algorithm that learns the value of each action in different states through continuous interaction with the environment. During training, whenever the Bluetooth headset is in a certain state, it selects an adjustment action to execute based on the value of each action in the current state. After executing the action, the reward value corresponding to that action is calculated using a reward function, and the next state is observed. Subsequently, the value of each action in different states is adjusted according to the value update formula for the current state and the corresponding action in different states. The adjustment process comprehensively considers the current reward value and the optimal value of each action in different states among all available actions in the next state, allowing the value of each action in different states to gradually converge towards the direction that yields higher rewards. Through repeated iterations of selecting actions, obtaining rewards, and updating the value of each action in different states, the decision-making strategy is continuously optimized, ultimately learning adjustment actions that yield higher rewards in different states—that is, a more effective anti-interference adjustment scheme.

[0078] The current state refers to the combined state of the interference environment and headphone performance that triggered this adjustment action, defined by multiple dimensions of parameters. Its composition must strictly follow the previously determined state space definition, which consists of the comprehensive interference level coefficient, the current signal-to-noise ratio, the current packet loss rate, and the current data transmission delay.

[0079] The different states in each action corresponding to different states refer to other states that are related to the current state in terms of scene. These states are usually similar to or related to the current state in terms of interference characteristics or headphone performance. For example, if the current state has a comprehensive interference level coefficient of 0.4 (moderate interference), a current signal-to-noise ratio of 26 dB, a current packet loss rate of 1.2%, and a current data transmission delay of 24 ms, its "corresponding different states" may include states with a comprehensive interference level coefficient of 0.35 (moderate interference), a current signal-to-noise ratio of 25 dB, a current packet loss rate of 1.5%, and a current data transmission delay of 25 ms, or states with a comprehensive interference level coefficient of 0.45 (moderate interference), a current signal-to-noise ratio of 27 dB, a current packet loss rate of 1.0%, and a current data transmission delay of 23 ms. The interference severity and headphone performance level of these states are similar to the current state, and they belong to the related scenes of the current state. The actions in different states refer to all the adjustment actions that the Bluetooth headset can perform in these different states. These adjustment actions belong to the same action space as the adjustment actions in the current state, that is, they all include the radio frequency gain adjustment and frequency hopping adjustment of the radio frequency module, the decoding algorithm mode switching and error hiding depth adjustment of the decoding module, and the antenna matching degree adjustment of the antenna module. It is a set of adjustable parameters covering all functional modules.

[0080] When the reward value is greater than or equal to 0.5, it indicates that the adjustment action performed in the current state has achieved excellent adjustment results—that is, the adjustment action can significantly improve the performance of the Bluetooth headset, specifically manifested in a significant increase in signal-to-noise ratio, a significant decrease in packet loss rate, and a significant reduction in data transmission latency after adjustment. At this point, the value weight of each action in the current state and its corresponding different states is increased to strengthen the priority of this effective adjustment action in the current state and its associated different states, ensuring that the Bluetooth headset will prioritize the verified effective adjustment action when encountering similar scenarios in the future. The specific adjustment logic includes two aspects: First, directly increasing the value weight of the effective adjustment action in the current state. That is, in the state and action value table, increasing the value value of the corresponding entries for the current state and the current adjustment action, so that the priority of the current adjustment action in the current state is higher than other unverified or poorly effective actions. Second, appropriately increasing the value weight of actions that are the same as or similar to the current effective adjustment action in different states. For example, if the action of "adjusting the RF gain to +2 dB and adjusting the frequency hopping frequency to 2.450 GHz" is effective in the current state, then the value weight of the action of "adjusting the RF gain to around +2 dB and adjusting the frequency hopping frequency to around 2.450 GHz" will also be appropriately increased in "different states," because such similar actions are likely to produce good adjustment effects in related scenarios as well. Through this weight adjustment, the priority of the current effective action can be consolidated, and the applicable scenarios of the effective action can be expanded to related states. This accelerates the convergence of strategy iteration towards "efficient actions covering more similar scenarios," ensuring that Bluetooth headphones can call effective adjustment actions more quickly in dynamic interference environments and continuously improve anti-interference effects.

[0081] Finally, the game relationship between the headphones and the interference source is analyzed using Nash equilibrium theory to determine the optimal set of adjustment actions. In a real interference environment, the interference source continuously affects the headphone performance; the two can be considered as two players in a game, where the headphone aims to maximize its own performance, and the interference source aims to minimize the headphone performance. To find a robust adjustment strategy, the payoff matrix under different adjustment actions needs to be calculated first. Each element in the payoff matrix corresponds to the headphone's performance value under a certain adjustment action, which is determined by the adjusted signal-to-noise ratio, packet loss rate, and data transmission delay. Subsequently, the Nash equilibrium point is searched in the payoff matrix. The action corresponding to this equilibrium point satisfies the following condition: under this action, even if the interference source adopts the most unfavorable strategy for the headphone, the headphone cannot further improve its performance by unilaterally changing the adjustment action. Including the action corresponding to the Nash equilibrium point in the optimal set of adjustment actions ensures that the adjustment actions in this set can maintain a stable anti-interference effect under the dynamic influence of the interference source, avoiding a significant decrease in adjustment effect due to changes in the interference source's strategy, and ultimately providing a Bluetooth headphone with an adjustment scheme that is both effective and robust.

[0082] The process involves using a state space comprised of a comprehensive interference level coefficient, the current signal-to-noise ratio, the current packet loss rate, and the current data transmission delay; an action space comprised of adjustable parameters of the Bluetooth headset; and a reward function constructed based on the performance improvement after adjustment to determine the optimal set of adjustment actions. This includes:

[0083] The process of defining a spatial state is as follows:

[0084] ;

[0085] Where S represents the spatial state, SNR represents the current signal-to-noise ratio, and L represents the packet loss rate. Current data transmission delay;

[0086] The process of defining the reward function is as follows:

[0087] ;

[0088] in, ; ; , The difference between the adjusted signal-to-noise ratio and the original signal-to-noise ratio. The difference between the adjusted packet loss rate and the original packet loss rate is used to calculate the packet loss rate. This is the difference between the adjusted data transmission delay and the original data transmission delay.

[0089] The optimization strategy process is as follows:

[0090] ;

[0091] in, Let S be the value of each action in different states corresponding to the current state S and action A. To update the value of each action in different states corresponding to action A before the update, For learning rate, As a discount factor, The next state after adjustment, This is an optional action for the next state. It represents the maximum value among all possible actions in the next state, corresponding to the values ​​of each action in different states.

[0092] S40: The adjustment parameters corresponding to the optimal set of adjustment actions are sent to the corresponding functional module of the Bluetooth headset. After continuous monitoring for a preset number of cycles, the average value is taken as the adjusted performance.

[0093] Specifically, parameter adjustment execution and closed-loop feedback optimization are key steps in implementing adaptive adjustment decisions and continuously iterating and optimizing the adjustment strategy. The core of this process is to ensure that the Bluetooth headset's anti-interference adjustment effect continuously adapts to the dynamic interference environment through a cycle of parameter execution, performance monitoring, and strategy updates. The specific process is as follows:

[0094] First, the adjustment parameters corresponding to the optimal set of adjustment actions are sent to the corresponding functional modules of the Bluetooth headset for execution. The optimal set of adjustment actions includes specific adjustment parameters for the RF module, decoding module, and antenna module. These parameters must precisely match the control logic of each functional module. Specifically, after receiving the adjustment parameters, the RF module adjusts the RF gain and frequency hopping frequency according to the parameter requirements. This increases or decreases the signal reception strength by changing the RF gain and avoids frequency bands with concentrated interference by switching the frequency hopping frequency. After receiving the adjustment parameters, the decoding module switches the decoding algorithm mode and adjusts the error concealment depth according to the parameter settings. It switches the decoding algorithm mode to an anti-interference mode more suitable for the current interference scenario and improves the compensation capability for lost data by adjusting the error concealment depth. After receiving the adjustment parameters, the antenna module adjusts the antenna matching degree according to the parameter values, optimizing the adaptation efficiency between the antenna and the signal transmission path and reducing energy loss during signal reception. All functional modules execute the adjustment parameters synchronously to ensure that the adjustment actions are implemented quickly and respond promptly to current interference.

[0095] Secondly, after the adjustment is performed, the performance indicators of the Bluetooth headset are collected at a preset monitoring period to determine the adjusted performance. The preset monitoring period needs to balance monitoring timeliness and data stability, ensuring rapid capture of performance changes after adjustment while avoiding data deviations caused by short-term fluctuations. According to this preset monitoring period, the performance indicators of the Bluetooth headset are continuously collected. These indicators specifically include the adjusted signal-to-noise ratio, adjusted packet loss rate, and adjusted data transmission delay. These three indicators reflect the adjusted headset performance from three dimensions: signal purity, data transmission integrity, and data transmission timeliness, respectively. To further reduce the impact of data fluctuations on performance evaluation, monitoring is continuously conducted for a preset number of periods. After completing the preset number of monitoring periods, the arithmetic mean of the collected adjusted signal-to-noise ratio, adjusted packet loss rate, and adjusted data transmission delay is calculated. This average is used as the final adjusted performance to ensure the accuracy of the performance evaluation results.

[0096] Finally, a reward value is calculated based on the difference between the adjusted performance and the performance before adjustment. This reward value is then used to update the state and action value tables, enabling strategy iteration and forming a closed-loop feedback optimization. The performance before adjustment refers to the Bluetooth headset performance data collected before the adjustment action, including the signal-to-noise ratio (SNR), packet loss rate, and data transmission latency before adjustment. The performance change is obtained by calculating the difference between the adjusted and pre-adjustment performance, namely the difference between the SNR, packet loss rate, and data transmission latency. These differences are then substituted into the reward function to calculate the reward value corresponding to this adjustment action. The magnitude of the reward value directly reflects the effectiveness of the adjustment action. Based on different reward value ranges, the values ​​of each action in the state and action value table under different states of the current state and corresponding adjustment action are updated differentially: When the reward value is greater than or equal to 0.5, it indicates excellent adjustment effect, and the value weight of each action under different states of the current state and corresponding action should be increased so that this action is selected first when encountering the same state in the future; when the reward value is greater than 0 and less than 0.5, it indicates moderate adjustment effect, and the value of each action under different states of the current state and corresponding action is updated normally, retaining this action as an alternative adjustment scheme; when the reward value is less than or equal to 0, it indicates ineffective adjustment or negative effect, and the value weight of each action under different states of the current state and corresponding action should be reduced, while triggering action mutation, randomly selecting an untried adjustment action to explore new effective adjustment directions. Each time adjustment parameter execution and performance monitoring is completed, that is, a process of parameter execution, performance monitoring, reward calculation, and value update of each action under different states is completed, the state and action value table is updated once, so that the decision strategy iterates according to the latest adjustment effect. As the number of iterations increases, the value of each action under different states in the state and action value table gradually converges towards a better direction. The adjustment strategy continuously adapts to the current interference environment, eventually forming a continuously optimized closed-loop feedback mechanism to ensure that the anti-interference capability of the Bluetooth headset always matches the dynamic interference environment.

[0097] S50: Calculate the reward value based on the difference between the adjusted performance and the pre-adjustment performance, update the state and action value table according to the reward value, realize strategy iteration, and form a closed-loop feedback optimization.

[0098] Specifically, sending the adjustment parameters corresponding to the optimal set of adjustment actions to the corresponding functional modules of the Bluetooth headset for execution is a crucial step in transforming the anti-interference scheme determined in the decision-making stage into actual operation. The optimal set of adjustment actions contains specific adjustment parameters for different functional modules of the Bluetooth headset. These parameters need to be precisely matched with the control logic and operating characteristics of each functional module. Specifically, after receiving the corresponding adjustment parameters, the RF module adjusts the RF gain and frequency hopping frequency according to the parameter settings. By changing the RF gain, it adapts to the signal reception requirements under the current interference intensity. By switching the frequency hopping frequency, it avoids the frequency bands where interference signals are concentrated, reducing the impact of interference on signal reception. After receiving the adjustment parameters, the decoding module switches the decoding algorithm mode and adjusts the error concealment depth according to the parameter requirements. It switches the decoding algorithm mode from standard mode to anti-interference mode to improve the decoding capability of audio data under interference conditions. By adjusting the error concealment depth, it enhances the compensation effect for lost data packets during transmission, ensuring the continuity of audio output. After receiving the adjustment parameters, the antenna module adjusts the antenna matching degree according to the parameter values, optimizing the adaptation relationship between the antenna and the signal transmission path, reducing energy loss during signal reception, and improving the reception efficiency of effective signals. Upon receiving the adjustment parameters, each functional module synchronously executes the adjustment operation to ensure that the adjustment action is implemented quickly and responds promptly to the current interference environment.

[0099] After adjustment, performance metrics of the Bluetooth headset are collected at a preset monitoring period. This preset monitoring period must balance the timeliness of performance changes with the stability of data collection, avoiding missing dynamic performance changes due to excessively long monitoring intervals or causing excessive data fluctuations due to excessively short intervals. Performance metrics of the Bluetooth headset are continuously collected according to this preset monitoring period. These metrics specifically include the adjusted signal-to-noise ratio (SNR), adjusted packet loss rate, and adjusted data transmission delay. The adjusted SNR reflects the ratio of effective signal to noise in the received signal after adjustment; the adjusted packet loss rate reflects the proportion of data packets lost during data transmission after adjustment; and the adjusted data transmission delay characterizes the time required for the adjusted data to be transmitted from the sending end to the receiving end. These three metrics together constitute the core dimensions for evaluating the adjustment effect. To further reduce the interference of short-term data fluctuations on the performance evaluation results, continuous monitoring is required for a preset number of periods. After completing the preset number of monitoring periods, the arithmetic mean of the collected adjusted SNR, adjusted packet loss rate, and adjusted data transmission delay is calculated. This average is used as the final adjusted performance, ensuring that the performance evaluation results accurately reflect the actual effect of the adjustment action.

[0100] Calculating the reward value based on the difference between the performance before and after adjustment is the core step in quantifying the effectiveness of the adjustment action. The performance before adjustment is Bluetooth headset performance data acquired according to the same standards and periods as the performance after adjustment, specifically including the signal-to-noise ratio (SNR), packet loss rate, and data transmission latency before adjustment. By calculating the difference between the performance before and after adjustment, the changes in each performance indicator are obtained: the difference in SNR, packet loss rate, and data transmission latency. These changes are then substituted into a preset reward function to calculate the reward value corresponding to this adjustment action. The magnitude of the reward value directly corresponds to the effect of the adjustment action on the headset performance. Based on different reward value ranges, the values ​​of each action in the state and action value table under different states of the current state and the corresponding adjustment action are updated differentially: When the reward value is greater than or equal to 0.5, it indicates that the adjustment action is excellent, and the value weight of each action under different states of the current state and the corresponding adjustment action should be increased so that the Bluetooth headset will prioritize this adjustment action when encountering the same state in the future; when the reward value is greater than 0 and less than 0.5, it indicates that the adjustment action is average, and the value of each action under different states of the current state and the corresponding adjustment action should be updated normally, and this action should be retained as an alternative adjustment scheme for use in other relevant states in the future; when the reward value is less than or equal to 0, it indicates that the adjustment action is ineffective or even has a negative impact on the headset performance, and the value weight of each action under different states of the current state and the corresponding adjustment action should be reduced. At the same time, the action mutation mechanism should be triggered to randomly select an untried adjustment action to explore new effective adjustment directions and avoid the strategy from getting stuck in local optima.

[0101] Each iteration of the process—including parameter execution, performance metric collection, reward calculation, and value updates for actions under different states—concludes a policy iteration. As the number of iterations increases, the values ​​of each state and its corresponding action in the state-action value table gradually converge towards a better outcome. The adjustment strategy continuously adapts to changes in the current interference environment, ultimately forming a closed-loop feedback optimization mechanism encompassing parameter execution, performance monitoring, reward calculation, value updates for actions under different states, and strategy optimization. This ensures that the Bluetooth headset's anti-interference capability always matches the dynamically changing interference environment, continuously improving its anti-interference performance.

[0102] In one embodiment, such as Figure 2 As shown, a Bluetooth headset adaptive anti-interference adjustment system is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the Bluetooth headset adaptive anti-interference adjustment method.

[0103] For specific limitations regarding electronic devices, please refer to the limitations on the adaptive anti-interference adjustment method for Bluetooth headsets mentioned above, which will not be repeated here.

[0104] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0105] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A Bluetooth headset adaptive anti-interference adjustment method, characterized in that, include: Intensity features, frequency features, and complexity features are extracted from the preprocessed signal, and each extracted feature is normalized to obtain normalized features. Based on the normalized features, a single-dimensional interference coefficient is calculated to obtain a comprehensive interference level coefficient. The interference level is then classified using the comparison result between the comprehensive interference level coefficient and a preset threshold. The interference level is used to provide the decision-making model with interference level classification and strategy optimization of the state and action value table. The state space is constructed using the comprehensive interference level coefficient, the current signal-to-noise ratio, the current packet loss rate, and the current data transmission delay. The action space is constructed using the adjustable parameters of the Bluetooth headset. The reward function is constructed using the performance improvement effect after adjustment to determine the optimal set of adjustment actions. The adjustment parameters corresponding to the optimal set of adjustment actions are sent to the corresponding functional module of the Bluetooth headset for continuous monitoring for a preset number of cycles, and the average value is taken as the performance after adjustment. The reward value is calculated based on the difference between the adjusted performance and the performance before adjustment. The state and action value table is updated according to the reward value to realize strategy iteration and form a closed-loop feedback optimization.

2. The Bluetooth headset adaptive anti-interference adjustment method according to claim 1, characterized in that, The preprocessed signal is used to extract intensity features, frequency features, and complexity features, including: The Bluetooth headset uses its built-in radio frequency receiver to collect interference signals from the surrounding environment in real time at a preset collection period to obtain the original signal sequence. The original signal sequence is subjected to filtering and denoising preprocessing to obtain the preprocessed signal; The preprocessed signal is subjected to a Fast Fourier Transform to calculate the main interference frequency and frequency fluctuation range, thereby extracting intensity features, frequency features, and complexity features. The intensity features include peak intensity and average intensity, the frequency features include the main interference frequency and frequency fluctuation range, and the complexity features include fractal dimension and Hearst exponent. The main interference frequency is the frequency corresponding to the maximum value of the frequency domain signal amplitude spectrum, and the frequency fluctuation range is the difference between the maximum and minimum frequencies of the frequency domain signal.

3. The Bluetooth headset adaptive anti-interference adjustment method according to claim 1, characterized in that, The single-dimensional interference coefficient is calculated based on the normalized features, including: Calculate the intensity anomaly coefficient: ； in, ; The The intensity anomaly coefficient, This is the sum of the energies of all high-frequency components after performing a discrete wavelet transform on the preprocessed signal. This is the sum of the total energy of the high-frequency components and the energy of the low-frequency components of the preprocessed signal after discrete wavelet transform. The sum of the energies of all low-frequency components after discrete wavelet transform of the preprocessed signal is given by J, where J is the preset total number of wavelet decomposition levels, and j is any wavelet decomposition level. The energy of the high-frequency component of the j-th layer; The frequency anomaly coefficient is calculated as follows: ； in, ; For frequency anomaly coefficients, The standard deviation of the frequency component amplitude, Let M be the average amplitude of the frequency components, M be the number of frequency points, and k be any frequency point. The frequency domain signal value at the k-th frequency point; Calculate the duration coefficient: ； in, Here, T is the duration coefficient, where T is the duration of the current interference. This is the short-term interference threshold. This is the long-term interference threshold; The formula for calculating the complexity coefficient is: ； in, is the complexity coefficient, D is the fractal dimension, and H is the Hearst exponent.

4. The Bluetooth headset adaptive anti-interference adjustment method according to claim 3, characterized in that, The formula for calculating the comprehensive interference level coefficient is as follows: ； in, The comprehensive interference level coefficient is denoted by 'a', where 'a' represents the weight of the intensity anomaly coefficient. Let be the intensity anomaly coefficient, and b be the weight of the frequency anomaly coefficient. Here, c represents the frequency anomaly coefficient, and c represents the weight of the duration coefficient. The duration coefficient and the weight of the complexity coefficient d are the following: is the complexity coefficient.

5. The Bluetooth headset adaptive anti-interference adjustment method according to claim 1, characterized in that, The state space is constructed using the comprehensive interference level coefficient, the current signal-to-noise ratio, the current packet loss rate, and the current data transmission delay; the action space is constructed using the adjustable parameters of the Bluetooth headset; and a reward function is constructed using the performance improvement effect after adjustment to determine the optimal set of adjustment actions, including: The process of defining a spatial state is as follows: ； Where S represents the spatial state, SNR represents the current signal-to-noise ratio, and L represents the packet loss rate. Current data transmission delay; The process of defining the reward function is as follows: ； in, ; ; , The difference between the adjusted signal-to-noise ratio and the original signal-to-noise ratio. The difference between the adjusted packet loss rate and the original packet loss rate is used to calculate the packet loss rate. This is the difference between the adjusted data transmission delay and the original data transmission delay. The optimization strategy process is as follows: ； in, Let S be the value of each action in different states corresponding to the current state S and action A. To update the value of each action in different states corresponding to action A before the update, For learning rate, As a discount factor, The next state after adjustment, This is an optional action for the next state. It represents the maximum value among all possible actions in the next state, corresponding to the values ​​of each action in different states.

6. The Bluetooth headset adaptive anti-interference adjustment method according to claim 1, characterized in that, The process of determining the optimal set of adjustment actions specifically includes: treating the headphones and the interference source as two parties in a game, with the headphones aiming to maximize performance and the interference source aiming to minimize headphone performance; calculating the payoff matrix under different actions, where each element in the payoff matrix represents the headphone's performance value under the corresponding action; finding the Nash equilibrium point in the payoff matrix, where the Nash equilibrium point represents an action that the headphones cannot further improve by unilaterally changing their actions; and selecting the action corresponding to the Nash equilibrium point as the preferred action and incorporating it into the optimal set of adjustment actions.

7. The Bluetooth headset adaptive anti-interference adjustment method according to claim 1, characterized in that, The specific process of updating the state and action value table based on reward value to achieve policy iteration includes: when the reward value is greater than or equal to 0.5, increasing the value weight of each action in different states corresponding to the current state and action; when the reward value is greater than 0 and less than 0.5, updating the value of each action in different states corresponding to the current state and action normally; when the reward value is less than or equal to 0, decreasing the value weight of each action in different states corresponding to the current state and action, and triggering action mutation, randomly selecting untried actions; updating the state and action value table once after each parameter adjustment execution and performance monitoring, thus achieving policy iteration.

8. The Bluetooth headset adaptive anti-interference adjustment method according to claim 1, characterized in that, The optimal set of adjustment actions was determined by analyzing the game relationship between the headphones and the interference source using reinforcement learning combined with Nash equilibrium theory.

9. The Bluetooth headset adaptive anti-interference adjustment method according to claim 1, characterized in that, The specific process of sending the adjustment parameters to the corresponding functional modules of the Bluetooth headset for execution and collecting the Bluetooth headset performance indicators includes: The adjustment parameters are sent to the RF module, decoding module, and antenna module. The RF module adjusts the RF gain and frequency hopping frequency according to the adjustment parameters. The decoding module switches the decoding algorithm mode and adjusts the error concealment depth according to the adjustment parameters. The antenna module adjusts the antenna matching degree according to the adjustment parameters.

10. A Bluetooth headset adaptive anti-interference adjustment system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the Bluetooth headset adaptive anti-interference adjustment method according to any one of claims 1 to 9.

Images