Intelligent speaker control system

CN118645101BActive Publication Date: 2024-12-03SHENZHEN ZHONGKE RUI TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411096568.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-12
Publication Date
2024-12-03
Estimated Expiration
2044-08-12

AI Technical Summary

Technical Problem

The existing smart speaker control system has shortcomings in speech recognition accuracy, resource management efficiency and environmental noise processing. It is difficult to accurately capture and process user voice commands in complex or noisy environments, which affects user experience and response speed, and limits voice control. The utility of technology in emergency application scenarios.

Method used

The autoregressive integral sliding average model and long and short-term memory network are used to analyze the voice mode, and the resource configuration is dynamically adjusted, combined with the Kalman filter for noise suppression, monitoring the response time and processing load to optimize the response speed to form an intelligent speaker control system.

Benefits of technology

It improves the recognition accuracy and execution accuracy of voice commands, improves user interaction experience, optimizes resource usage efficiency, ensures accurate identification of user commands in noisy environments, improves system response speed and processing capabilities, and saves energy consumption.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118645101B_ABST
    Figure CN118645101B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of voice control technology, specifically to an intelligent speaker control system, the system includes a voice data analysis module, a dynamic resource management module, a noise suppression adjustment module, and an intelligent response speed control module. In the present invention, by accurately analyzing the voice input of the speaker, the recognition accuracy and execution effect of the voice command are improved, the user interaction experience is significantly improved, the autoregressive integral sliding average model and the long short-term memory network are used to enhance the processing capability and data efficiency of complex commands, implement dynamic resource management, automatically adjust the CPU and memory configuration according to command prediction, optimize the device response speed and resource utilization, improve the system processing capacity, and ensure that the real-time noise suppression can accurately capture the command in a multi-noise environment. In addition, by continuously monitoring and optimizing the response speed and load of the intelligent speaker, the operation efficiency is improved, while reducing energy consumption and processing burden.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of voice control technology, and in particular to an intelligent speaker control system. Background Art

[0002] The field of voice control technology involves the operation and control of devices through sound and voice commands. This technology uses a voice recognition system to parse the user's verbal commands and convert them into executable commands to control various devices and services. This technical field includes voice-to-text conversion, command recognition, natural language processing (NLP), and the development of machine learning algorithms to improve recognition accuracy and the ability to understand complex commands. Voice control technology has been widely used in home automation, in-vehicle systems, smartphone applications, and assistive technologies. Its core advantage is that it provides a contactless operation method, enhances the user experience, and makes device operation more intuitive and convenient.

[0003] Among them, the smart speaker control system is a system that uses voice control technology to operate smart speakers. A smart speaker is a device equipped with a microphone, a speaker and an Internet connection that can respond to the user's voice commands to perform various functions, such as playing music, setting alarms, providing weather forecasts, controlling smart home devices, etc. Through the smart speaker control system, users can communicate with the device only through voice interaction without manual operation, making daily life more convenient and intelligent. The purpose of this system is mainly to simplify device management in a home or office environment through a natural language interface and improve the level of automation in life.

[0004] Although the existing smart speaker control system is widely used in a variety of speaker devices, it still has deficiencies in voice recognition accuracy, resource management efficiency and environmental noise processing. Especially in complex or noisy environments, the existing technology is difficult to accurately capture and process user voice commands due to environmental noise. The traditional voice recognition system lacks flexibility in resource allocation, resulting in poor operation on devices with limited resources, affecting the user experience. The optimization of response speed has also failed to meet the requirements of real-time updates, which limits the effectiveness of voice control technology in emergency or rapid response application scenarios, which not only affects the user's operational convenience, but also limits the application and promotion of voice technology in a wider range of scenarios. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an intelligent speaker control system.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: an intelligent speaker control system comprises:

[0007] The voice data analysis module analyzes the input data of the speaker through the autoregressive integral sliding average model to obtain the voice pattern analysis results, and uses the long short-term memory network to refine the prediction of the command pattern to obtain the refined command prediction results;

[0008] The dynamic resource management module dynamically adjusts the resources controlled by the smart speaker according to the refined command prediction results, including the CPU allocation and memory configuration, performs resource optimization, obtains the speaker resource optimization status, adjusts the operating parameters according to the speaker resource optimization status, and obtains the optimized resource configuration;

[0009] The noise suppression adjustment module captures the ambient audio data based on the optimized resource configuration, analyzes and updates the state of the audio signal through the Kalman filter, obtains the audio state adjustment result, uses the audio state adjustment result to suppress the ambient noise in real time, captures the voice command controlled by the smart speaker, and obtains the optimized voice capture result;

[0010] The intelligent response speed control module monitors the response time and processing load of the smart speaker based on the optimized voice capture result, generates a speaker performance log, analyzes the response efficiency and load according to the speaker performance log, optimizes the response speed, and obtains the adjusted speaker performance control response configuration.

[0011] As a further solution of the present invention, the step of obtaining the refined command prediction result is specifically as follows:

[0012] Based on the speaker voice input data, spectrum analysis is applied for signal processing to extract the fundamental frequency and spectral features using the formula:

[0013] ;

[0014] Calculate the fundamental frequency characteristics and generate signal characteristic analysis results, where: represents the fundamental frequency analysis result, is the ith frequency component, is the corresponding amplitude, is the number of sampling points, It is a continuous indicator;

[0015] Using the signal feature analysis results, a long short-term memory network is used to learn and predict the command mode using the formula:

[0016] ;

[0017] Strengthen the model's sensitivity to input variation, calculate command pattern probabilities, and generate preliminary command pattern prediction results, where is the predicted probability of the command mode, is the weight parameter, is the output function of the LSTM unit, are input features, is the number of features, is the coefficient of variation;

[0018] The results of the preliminary command pattern prediction are used, combined with the context information of the current environment, using the formula:

[0019] ;

[0020] Calculate and obtain the refined command prediction results, where: is the command prediction result, is the probability of the jth predicted command, is the number of commands referenced, is the standard deviation.

[0021] As a further solution of the present invention, the step of obtaining the speaker resource optimization state is specifically as follows:

[0022] The resource demand sensitivity parameter of the command is extracted from the refined command prediction result, and the resource demand of the command is calculated using the formula:

[0023] ;

[0024] Generate detailed command resource requirements, where: Represents resource demand, represents the resource sensitivity of the command prediction, is the adjustment coefficient;

[0025] According to the detailed command resource requirements, the CPU and memory configuration of the smart speaker are adjusted using the formula:

[0026] ;

[0027] Generate an updated resource allocation, where: Indicates the updated resource allocation. is the adjustment parameter;

[0028] Based on the updated resource allocation, monitor and record the running status after resource optimization, using the formula:

[0029] ;

[0030] Generates the speaker resource optimization state, where It is the speaker resource optimization state. is a performance evaluation parameter.

[0031] As a further solution of the present invention, the step of obtaining the optimized resource configuration is specifically:

[0032] According to the speaker resource optimization state, the formula is used:

[0033] ;

[0034] Generate the optimized state parameters after analysis, where represents the optimized state parameters after analysis, is the gain factor, is the baseline correction parameter;

[0035] According to the optimized state parameters after the analysis, the required CPU and memory resources are calculated using the formula:

[0036] ;

[0037] Generates the evaluated resource parameters, where Represents the resource parameters after evaluation, is the adjustment factor;

[0038] Using the resource parameters evaluated above, the operating parameters are adjusted to optimize the sound control performance using the formula:

[0039] ;

[0040] Generate an optimized resource configuration, where: It is the optimized resource allocation. is the weight parameter, is to adjust the proportion, Represents the current resource configuration.

[0041] As a further solution of the present invention, the step of obtaining the audio state adjustment result is specifically:

[0042] Based on the optimized resource configuration, audio capture is activated to collect environmental audio data using the formula:

[0043] ;

[0044] Generate captured audio data parameters, where Represents the captured audio data, and are the gain and sensitivity adjustment factors, is the ambient audio intensity;

[0045] The captured audio data parameters are input into the Kalman filter to analyze and update the state of the audio signal, and adjust the smoothness and response speed of the signal, using the formula:

[0046] ;

[0047] Generate a filtered audio signal where represents the filtered audio signal, and It is to adjust the sensitivity and threshold of signal processing;

[0048] Based on the filtered audio signal, the audio state is updated using the formula:

[0049] ;

[0050] Generate an audio state adjustment result, where: is the result of audio status adjustment. and Adjust the update rate and adjustment range of the audio status.

[0051] As a further solution of the present invention, the step of obtaining the optimized voice capture result is specifically:

[0052] According to the audio state adjustment result, real-time noise suppression is performed to optimize the audio capture in the environment, using the formula:

[0053] ;

[0054] Generate noise suppressed audio parameters, where is the optimized noise suppression parameter, They are noise suppression gain, stability coefficient and sensitivity adjustment parameters respectively;

[0055] The audio parameters after noise suppression are used to capture and preliminarily process the voice commands controlled by the smart speaker, using the formula:

[0056] ;

[0057] Generate preliminary processed speech parameters, where is the voice command parameter after preliminary processing, and are the processing gain and weight parameters, The original voice data in the environment;

[0058] According to the initially processed speech parameters, the speech recognition configuration of the smart speaker is updated using the formula:

[0059] ;

[0060] Generate optimized speech capture results, where For the optimized voice capture results, represents the update coefficient.

[0061] As a further solution of the present invention, the steps of obtaining the speaker performance log are specifically as follows:

[0062] Using the optimized voice capture results, the characterization of the response time is optimized, using the formula:

[0063] ;

[0064] Generate response time performance indicators, where and represents the gain factor, basic bias and adjustment factor, Represents the response time performance indicator;

[0065] Combining the response time performance indicator and processing load, the formula is used:

[0066] ;

[0067] Generate performance log parameters, where: is the performance log parameter, and is the weight parameter, Indicates processing load;

[0068] Using the performance log parameters, by configuring the speaker parameters and recording key performance data, the formula is:

[0069] ;

[0070] Generates a speaker performance log, where is the fusion coefficient, This is the previous performance log. It is a log of speaker performance.

[0071] As a further solution of the present invention, the step of obtaining the adjusted speaker performance control response configuration is specifically:

[0072] Extract key performance indicators from the speaker performance log, including response efficiency and processing load, and calculate preliminary performance data using the formula:

[0073] ;

[0074] Generate performance analysis results, where Represents performance data extracted from logs, , Weight coefficient, Represents the processing load, Indicates response efficiency;

[0075] Based on the performance analysis results, adjust the processing parameters of the speaker to optimize the response speed of the device, through the formula:

[0076] ;

[0077] Generate an optimized processing configuration where: For the optimized processing configuration, and is the adjustment parameter, Indicates adjusting the processing parameters of the speaker;

[0078] Using the optimized processing configuration, update the speaker's performance control configuration and apply the formula:

[0079] ;

[0080] The adjusted speaker performance control response configuration is obtained, wherein: is the update coefficient, It is the previous speaker performance control configuration. Indicates the adjusted speaker performance control response configuration.

[0081] Compared with the prior art, the advantages and positive effects of the present invention are:

[0082] In the present invention, the recognition accuracy and execution precision of voice commands are improved through careful analysis and processing of the speaker voice input, so that the user interaction experience is significantly improved. The autoregressive integral sliding average model and long short-term memory network are used to analyze the voice mode, which not only optimizes the data processing efficiency, but also improves the ability to understand complex commands. The implementation of dynamic resource management adjusts the CPU and memory configuration according to the prediction results, so that the device can automatically optimize resource usage under different requirements and improve the overall response speed and processing capacity of the system. Through real-time noise suppression, the quality of voice capture is further improved to ensure that the user's instructions can be accurately recognized even in a noisy environment. The monitoring and optimization of the response speed and processing load of the smart speaker can keep the device running efficiently while saving energy consumption and reducing unnecessary processing burden. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 is a system flow chart of the present invention;

[0084] Figure 2 It is a flow chart of the refined command prediction result in the present invention;

[0085] Figure 3 A flow chart showing the speaker resource optimization state in the present invention;

[0086] Figure 4 It is a flow chart of optimized resource configuration in the present invention;

[0087] Figure 5 A flowchart of the audio state adjustment result in the present invention;

[0088] Figure 6 This is a flow chart of the optimized voice capture result in the present invention;

[0089] Figure 7 A flow chart of the speaker performance log in the present invention;

[0090] Figure 8 A flow chart of the adjusted speaker performance control response configuration in the present invention. DETAILED DESCRIPTION

[0091] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0092] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0093] Example: See Figure 1 , an intelligent speaker control system comprises:

[0094] The voice data analysis module analyzes the input data of the speaker through the autoregressive integral sliding average model to obtain the voice pattern analysis results, and uses the long short-term memory network to refine the prediction of the command pattern to obtain the refined command prediction results;

[0095] The dynamic resource management module dynamically adjusts the resources controlled by the smart speaker according to the refined command prediction results, including CPU allocation and memory configuration, performs resource optimization, obtains the speaker resource optimization status, adjusts the operating parameters according to the speaker resource optimization status, and obtains the optimized resource configuration;

[0096] The noise suppression adjustment module captures the ambient audio data based on the optimized resource configuration, analyzes and updates the state of the audio signal through the Kalman filter, obtains the audio state adjustment result, uses the audio state adjustment result to suppress the ambient noise in real time, captures the voice commands controlled by the smart speaker, and obtains the optimized voice capture result;

[0097] The intelligent response speed control module monitors the response time and processing load of the smart speaker based on the optimized voice capture results, generates speaker performance logs, analyzes the response efficiency and load conditions according to the speaker performance logs, optimizes the response speed, and obtains the adjusted speaker performance control response configuration.

[0098] The results of voice pattern analysis include pitch distribution, sound spectrum, and speaking speed range. The refined command prediction results include command intent type, parameter setting range, and execution priority. The speaker resource optimization status includes processor load, memory occupancy, and energy consumption level. The optimized resource configuration includes resource allocation plan, storage optimization level, and processing power improvement. The audio status adjustment results include noise elimination effect, sound enhancement level, and signal stability. The optimized voice capture results include improved recognition accuracy, optimized response speed, and reduced error recognition rate. The speaker performance log includes performance peak records, response abnormality records, and resource usage history. The optimized speaker performance regulation response configuration includes time response optimization, load balancing strategy, and performance adaptive adjustment.

[0099] See also Figure 2 , the steps to obtain the command prediction results after refinement are as follows:

[0100] Based on the speaker voice input data, spectrum analysis is applied for signal processing to extract the fundamental frequency and spectral features using the formula:

[0101] ;

[0102] Calculate the fundamental frequency characteristics and generate signal characteristic analysis results, where: represents the fundamental frequency analysis result, is the ith frequency component, is the corresponding amplitude, is the number of sampling points, It is a continuous indicator;

[0103] Using the signal feature analysis results, the long short-term memory network is used to learn and predict the command mode using the formula:

[0104] ;

[0105] Strengthen the model's sensitivity to input variation, calculate command pattern probabilities, and generate preliminary command pattern prediction results, where is the predicted probability of the command mode, is the weight parameter, is the output function of the LSTM unit, are input features, is the number of features, is the coefficient of variation;

[0106] Use the preliminary command pattern to predict the result, combined with the contextual information of the current environment, and use the formula:

[0107] ;

[0108] Calculate and obtain the refined command prediction results, where: is the command prediction result, is the probability of the jth predicted command, is the number of commands referenced, is the standard deviation.

[0109] The formula for fundamental frequency analysis results is:

[0110] ;

[0111] : The i-th frequency component is generally obtained by converting the time domain signal into the frequency domain signal through Fourier transform;

[0112] : Corresponding frequency The amplitude of is obtained by Fourier transform result;

[0113] : The persistence index of the signal can be obtained by analyzing the length of time the signal is at a certain frequency;

[0114] : The number of sampling points, representing the total number of frequency components in the signal;

[0115] Assume the signal has three frequency components Hz, corresponding to the amplitude , the sustainability index of each component , the total number of sampling points ;

[0116] Calculation process :

[0117] ;

[0118] ;

[0119] this The value represents the weighted average fundamental frequency component in the signal and is a quantitative expression of the signal characteristics.

[0120] The predicted probability formula of command mode:

[0121] ;

[0122] : Weight parameter, which reflects the importance of each feature and is generally determined by statistical analysis of training data;

[0123] : The output function of the LSTM unit depends on the input features and the parameters of the network;

[0124] : Input features are preprocessed signal features such as MFCC;

[0125] : Coefficient of variation, indicating characteristics The variability in the training set can be obtained by statistical methods;

[0126] : number of features;

[0127] Assume there are three features, weights , , coefficient of variation , the number of features ;

[0128] calculate :

[0129] ;

[0130] ;

[0131] ;

[0132] this The values ​​represent the probability of the predicted command pattern after taking into account the variability.

[0133] Command prediction result formula:

[0134] ;

[0135] : The probability of the jth predicted command, from the above calculated;

[0136] : Standard deviation, indicating The dispersion of is obtained through statistical analysis;

[0137] : The number of commands considered;

[0138] Assume the prediction result , standard deviation , number of commands ;

[0139] calculate :

[0140] ;

[0141] ;

[0142] ;

[0143] ;

[0144] this The value represents a stability measure of the final command prediction results after adjustment for the comprehensive standard deviation.

[0145] See also Figure 3 , the specific steps for obtaining the speaker resource optimization status are:

[0146] Extract the resource demand sensitivity parameters of the command from the refined command prediction results and calculate the resource demand of the command using the formula:

[0147] ;

[0148] Generate detailed command resource requirements, where: Represents resource demand, represents the resource sensitivity of the command prediction, is the adjustment coefficient;

[0149] According to the detailed command resource requirements, adjust the CPU and memory configuration of the smart speaker using the formula:

[0150] ;

[0151] Generate an updated resource allocation, where: Indicates the updated resource allocation. Represents resource demand, is the adjustment parameter;

[0152] Based on the updated resource allocation, monitor and record the running status after resource optimization, using the formula:

[0153] ;

[0154] Generates the speaker resource optimization state, where It is the speaker resource optimization state. is the updated resource allocation, is a performance evaluation parameter.

[0155] Resource requirement formula:

[0156] ;

[0157] : Assume that it is the resource sensitivity parameter for command prediction, which is the model prediction value obtained by analyzing the user command type and frequency, such as ;

[0158] : Resource demand adjustment coefficient, assumed to be a fixed value, such as ;

[0159] calculate :

[0160] ;

[0161] calculate :

[0162] ;

[0163] Substituting the above results into :

[0164] ;

[0165] In this example, Indicates the resource demand intensity predicted by the command.

[0166] Updated resource allocation formula:

[0167] ;

[0168] : Obtained from the previous step, for example ;

[0169] :Adjust parameters, set ;

[0170] calculate :

[0171] ;

[0172] Calculate the denominator:

[0173] ;

[0174] Substituting the above results into :

[0175] ;

[0176] In this example, It indicates the optimized resource configuration state, which means that the resource allocation algorithm can be dynamically adjusted to improve efficiency.

[0177] Speaker resource optimization status formula:

[0178] ;

[0179] : Obtained from the previous step, for example ;

[0180] :Performance evaluation parameters, set ;

[0181] Calculate the second term:

[0182] ;

[0183] Substituting the above results into :

[0184] ;

[0185] In this example, It indicates the final resource optimization status of the speaker, reflects the overall benefit of resource optimization, and quantifies the actual status after performance optimization.

[0186] See also Figure 4 , the specific steps for obtaining the optimized resource configuration are:

[0187] According to the speaker resource optimization status, the formula is used:

[0188] ;

[0189] Generate the optimized state parameters after analysis, where represents the optimized state parameters after analysis, is the gain factor, is the baseline correction parameter, It is the speaker resource optimization status;

[0190] According to the analyzed optimization status parameters, the required CPU and memory resources are calculated using the formula:

[0191] ;

[0192] Generates the evaluated resource parameters, where Represents the resource parameters after evaluation, is the adjustment factor, Represents the optimized state parameters after analysis;

[0193] Use the estimated resource parameters to adjust the operating parameters and optimize the sound control performance using the formula:

[0194] ;

[0195] Generate an optimized resource configuration, where: It is the optimized resource allocation. is the weight parameter, is to adjust the proportion, and They represent the evaluated resource parameters and current resource configuration respectively.

[0196] The optimized state parameter formula after analysis:

[0197] ;

[0198] : Gain coefficient, assuming the value is 2, used to enhance the feedback effect of the state;

[0199] : Baseline correction parameter, assumed to be 1, used for equilibrium state response;

[0200] : Speaker resource optimization status, assuming the value is 3;

[0201] calculate :

[0202] ;

[0203] Calculate the denominator :

[0204] ;

[0205] Substituting the above results into formula:

[0206] ;

[0207] In this example, Represents the optimized state after analysis, indicating that the optimized state has been effectively amplified and adjusted.

[0208] The resource parameter formula after evaluation:

[0209] ;

[0210] : Adjustment factor, assumed to be 1.5, used to optimize the sensitivity of resource allocation;

[0211] : Adjustment factor, assuming the value is 0.5;

[0212] : The output obtained from the previous step, the value is 4.5;

[0213] calculate :

[0214] ;

[0215] Calculate the denominator :

[0216] ;

[0217] Substituting the above results into formula:

[0218] ;

[0219] In this example, Represents the resource demand after calculation, reflecting the high demand state for resource allocation.

[0220] Optimized resource allocation formula:

[0221] ;

[0222] : Weight parameter, assuming a value of 0.7, controls the balance between the new configuration and the current configuration;

[0223] : The adjustment ratio of the current configuration, assuming the value is 0.3;

[0224] : The output obtained from the previous step, the value is 28.0;

[0225] : Current resource configuration status, assumed to be 10;

[0226] calculate :

[0227] ;

[0228] calculate :

[0229] ;

[0230] Substituting the above results into formula:

[0231] ;

[0232] In this example, It indicates the optimized resource configuration and displays the actual adjustment results of the resource configuration.

[0233] See also Figure 5 , the specific steps for obtaining the audio status adjustment result are:

[0234] Based on the optimized resource configuration, activate audio capture and collect environmental audio data using the formula:

[0235] ;

[0236] Generate captured audio data parameters, where Represents the captured audio data, and are the gain and sensitivity adjustment factors, is the ambient audio intensity;

[0237] The captured audio data parameters are input into the Kalman filter to analyze and update the state of the audio signal, adjust the smoothness and response speed of the signal, and use the formula:

[0238] ;

[0239] Generate a filtered audio signal where represents the filtered audio signal, and is to adjust the sensitivity and threshold of signal processing, Represents the captured audio data;

[0240] Based on the filtered audio signal, update the audio state using the formula:

[0241] ;

[0242] Generate an audio state adjustment result, where: is the result of audio status adjustment. and Adjust the update rate and adjustment range of the audio status, Represents the filtered audio signal.

[0243] Captured audio data formula:

[0244] ;

[0245] : Gain factor, used to adjust the amplification level of the audio data, assuming ;

[0246] : Sensitivity adjustment factor, used to enhance the response to the ambient audio intensity, assuming ;

[0247] : Ambient audio intensity, assuming it is measured in a certain environment unit;

[0248] calculate :

[0249] ;

[0250] calculate :

[0251] ;

[0252] calculate :

[0253] ;

[0254] Substituting the above results into :

[0255] ;

[0256] In this example, Represents the result of processing the captured audio data, indicating the sensitivity and response to the ambient volume.

[0257] The filtered audio signal formula is:

[0258] ;

[0259] : Filter gain coefficient, used to adjust the smoothness of the audio signal, assuming ;

[0260] : Filter baseline parameters, ensure that the denominator is not zero, assuming ;

[0261] : Filter sensitivity adjustment parameter, affecting the signal response speed, assuming ;

[0262] : From the previous step, the value is 0.7455;

[0263] calculate :

[0264] ;

[0265] calculate :

[0266] ;

[0267] Calculate the denominator:

[0268] ;

[0269] Substituting the above results into :

[0270] ;

[0271] In this example, Represents the filtered audio signal, showing the effectiveness of the Kalman filter in reducing noise and smoothing the signal.

[0272] Audio status adjustment result formula:

[0273] ;

[0274] : State update coefficient, adjusts the update rate of the audio signal, assuming ;

[0275] : Amplitude adjustment parameter, used to simulate the periodic adjustment of the audio signal, assuming ;

[0276] : From the previous step, the value is 0.898;

[0277] Calculate sin :

[0278] ;

[0279] Substituting the above results into :

[0280] ;

[0281] In this example, It represents the final audio state adjustment result, showing the final processing effect of the audio signal achieved by the Kalman filter and the state update algorithm.

[0282] See also Figure 6 , the steps for obtaining the optimized voice capture results are as follows:

[0283] Adjust the results based on the audio state and perform real-time noise suppression to optimize audio capture in the environment using the formula:

[0284] ;

[0285] Generate noise suppressed audio parameters, where is the optimized noise suppression parameter, are noise suppression gain, stability coefficient and sensitivity adjustment parameters respectively, Indicates the audio status adjustment result;

[0286] The audio parameters after noise suppression are used to capture and preliminarily process the voice commands controlled by the smart speaker, using the formula:

[0287] ;

[0288] Generate preliminary processed speech parameters, where is the voice command parameter after preliminary processing, and are the processing gain and weight parameters, is the optimized noise suppression parameter, The original voice data in the environment;

[0289] According to the initially processed speech parameters, update the speech recognition configuration of the smart speaker using the formula:

[0290] ;

[0291] Generate optimized speech capture results, where For the optimized voice capture results, represents the update coefficient, It is the processed voice command parameter.

[0292] Optimized noise suppression parameter formula:

[0293] ;

[0294] (Noise Suppression Gain): Set to 0.5 to adjust the strength of noise suppression;

[0295] (Stability coefficient): Set to 1 to ensure that the denominator is not zero and balance noise suppression;

[0296] (Sensitivity adjustment parameter): Set to 0.1 to adjust the sensitivity of noise suppression;

[0297] (Audio state adjustment result obtained from the previous process): Assume it is 3;

[0298] calculate :

[0299] ;

[0300] calculate :

[0301] ;

[0302] Calculate the denominator:

[0303] ;

[0304] Substituting the above results into :

[0305] ;

[0306] In this example, Represents the audio parameters after noise suppression, indicating the effective noise suppression gain and stability adjustment.

[0307] The formula of voice command parameters after preliminary processing:

[0308] ;

[0309] (Processing Gain): Set to 2 to enhance the clarity of voice data;

[0310] (Weight parameter): Set to 0.8 to increase the weight of environmental data;

[0311] : Obtained from the previous step, the value is 2.59;

[0312] (Original voice data in the environment): Assume it is 5;

[0313] calculate :

[0314] ;

[0315] calculate :

[0316] ;

[0317] Substituting the above results into :

[0318] ;

[0319] In this example, Represents the speech parameters after preliminary processing, showing effective speech enhancement.

[0320] Optimized voice capture result formula:

[0321] ;

[0322] (Update coefficient): set to 0.7 to balance new and old data;

[0323] : Obtained from the previous step, the value is 4.86;

[0324] (Previous speech recognition result): Assume it is 3;

[0325] calculate and :

[0326] ;

[0327] ;

[0328] Adding the above results together we get :

[0329] ;

[0330] In this example, Represents the final optimized speech capture result, showing the effective integration of new and old data, ensuring the continuity and enhancement of speech recognition.

[0331] See also Figure 7 , the specific steps for obtaining the speaker performance log are:

[0332] Using the optimized speech capture results, the characterization of the response time is optimized, using the formula:

[0333] ;

[0334] Generate response time performance indicators, where and represents the gain factor, basic bias and adjustment factor, Represents the response time performance indicator, Indicates the optimized speech capture result;

[0335] Combining the response time performance indicator and the processing load, the formula is used:

[0336] ;

[0337] Generate performance log parameters, where: is the performance log parameter, and is the weight parameter, Represents the response time performance indicator, Indicates processing load;

[0338] Use the performance log parameters to configure the speaker parameters and record key performance data using the formula:

[0339] ;

[0340] Generates a speaker performance log, where is the fusion coefficient, This is the previous performance log. It is the speaker performance log. It is the performance log parameter.

[0341] Response time performance indicator formula:

[0342] ;

[0343] : Gain coefficient, set to 1.5 to enhance the impact of response time;

[0344] : Base bias, set to 2.0 to prevent the denominator from being zero;

[0345] : Adjustment coefficient, set to 0.5 to balance the contribution of response time to performance;

[0346] : The optimized speech capture result obtained from the previous processing step, assumed to be 4;

[0347] calculate :

[0348] ;

[0349] Calculate the denominator :

[0350] ;

[0351] Compute the full denominator:

[0352] ;

[0353] final value:

[0354] ;

[0355] This value Represents the response time performance indicator, revealing the impact of response time on the overall performance.

[0356] Performance log parameter formula:

[0357] ;

[0358] : Weight coefficient, set to 0.1, to adjust Contribution

[0359] : Adjustment coefficient, set to 1 to enhance The impact of

[0360] : The value calculated above is 6;

[0361] : Processing load, assuming it is 10;

[0362] calculate :

[0363] ;

[0364] calculate :

[0365] ;

[0366] final value:

[0367] ;

[0368] This value Represents a comprehensive performance log parameter that reflects the combined impact of processing load and response time.

[0369] Speaker performance log formula:

[0370] ;

[0371] : Fusion coefficient, set to 0.8, to balance the impact of new and old data;

[0372] : The value calculated above is 24.916;

[0373] : The previous performance log, assuming it is 20;

[0374] calculate :

[0375] ;

[0376] calculate :

[0377]

[0378] final value:

[0379] ;

[0380] This value Represents the final loudspeaker performance log, demonstrating the ability of performance monitoring to continuously update and integrate past data.

[0381] See also Figure 8 , the steps for obtaining the adjusted speaker performance control response configuration are as follows:

[0382] Extract key performance indicators from the speaker performance log, including response efficiency and processing load, and calculate preliminary performance data using the formula:

[0383] ;

[0384] Generate performance analysis results, where Represents performance data extracted from logs, , Weight coefficient, Represents the processing load, Indicates response efficiency;

[0385] Based on the performance analysis results, adjust the processing parameters of the speaker to optimize the response speed of the device, through the formula:

[0386] ;

[0387] Generate an optimized processing configuration where: For the optimized processing configuration, and is the adjustment parameter, Indicates adjusting the processing parameters of the speaker. Represents performance data extracted from logs;

[0388] Using the optimized processing configuration, update the speaker's performance control configuration and apply the formula:

[0389] ;

[0390] The adjusted speaker performance control response configuration is obtained, wherein: is the update coefficient, It is the previous speaker performance control configuration. For the optimized processing configuration, Indicates the adjusted speaker performance control response configuration.

[0391] Formula for extracting performance data from logs:

[0392] ;

[0393] Assume the parameters are:

[0394] : Weight coefficient, enhancing the impact of response efficiency;

[0395] : Weight coefficient, adjusting the sensitivity to processing load;

[0396] : Assume the response efficiency value is 5;

[0397] : Assume the processing load is 16;

[0398] calculate :

[0399] ;

[0400] calculate :

[0401] ;

[0402] Calculate the denominator :

[0403] ;

[0404] calculate :

[0405] ;

[0406] Final calculation :

[0407] ;

[0408] This shows , represents the weighted result of the performance data extracted from the logs.

[0409] Optimized processing configuration formula:

[0410] ;

[0411] Assume the parameters are:

[0412] : Adjust the square effect of performance data;

[0413] : Make sure the denominator is not zero;

[0414] : Flexibility in adjusting log parameters;

[0415] : Assume that the processing parameter is 3;

[0416] calculate :

[0417] ;

[0418] Calculating molecules :

[0419] ;

[0420] calculate :

[0421] ;

[0422] calculate :

[0423] ;

[0424] This shows , indicating that based on performance data The optimization result of the response speed configuration.

[0425] Adjusted speaker performance control response configuration formula:

[0426] ;

[0427] Assume the parameters are:

[0428] : The ratio of integrating new and old configurations;

[0429] : Assume that the previous configuration is 300;

[0430] calculate :

[0431] ;

[0432] calculate :

[0433] ;

[0434] calculate :

[0435] ;

[0436] This shows , represents the final updated speaker performance control response configuration, providing a comprehensive configuration based on previous and new optimization data.

[0437] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A smart speaker control system, It is characterized in that The system comprises: The voice data analysis module analyzes the input data of the speaker through the autoregressive integral sliding average model to obtain the voice pattern analysis results, and uses the long short-term memory network to refine the prediction of the command pattern to obtain the refined command prediction results; The dynamic resource management module dynamically adjusts the resources controlled by the smart speaker according to the refined command prediction results, including the CPU allocation and memory configuration, performs resource optimization, obtains the speaker resource optimization status, adjusts the operating parameters according to the speaker resource optimization status, and obtains the optimized resource configuration; The noise suppression adjustment module captures the ambient audio data based on the optimized resource configuration, analyzes and updates the state of the audio signal through the Kalman filter, obtains the audio state adjustment result, uses the audio state adjustment result to suppress the ambient noise in real time, captures the voice command controlled by the smart speaker, and obtains the optimized voice capture result; The intelligent response speed control module monitors the response time and processing load of the smart speaker based on the optimized voice capture result, generates a speaker performance log, analyzes the response efficiency and load according to the speaker performance log, optimizes the response speed, and obtains the adjusted speaker performance control response configuration.

2. The smart speaker control system according to claim 1, It is characterized in that The steps for obtaining the refined command prediction results are specifically as follows: Based on the speaker voice input data, spectrum analysis is applied for signal processing to extract the fundamental frequency and spectral features using the formula: ; Calculate the fundamental frequency characteristics and generate signal characteristic analysis results, where: represents the fundamental frequency analysis result, is the ith frequency component, is the corresponding amplitude, is the number of sampling points, It is a continuous indicator; Using the signal feature analysis results, a long short-term memory network is used to learn and predict the command mode using the formula: ; Strengthen the model's sensitivity to input variation, calculate command pattern probabilities, and generate preliminary command pattern prediction results, where is the predicted probability of the command mode, is the weight parameter, is the output function of the LSTM unit, are input features, is the number of features, is the coefficient of variation; The results of the preliminary command pattern prediction are used, combined with the context information of the current environment, using the formula: ; Calculate and obtain the refined command prediction results, where: is the command prediction result, is the probability of the jth predicted command, is the number of commands referenced, is the standard deviation.

3. The smart speaker control system according to claim 2, It is characterized in that The steps for obtaining the speaker resource optimization status are specifically as follows: The resource demand sensitivity parameter of the command is extracted from the refined command prediction result, and the resource demand of the command is calculated using the formula: ; Generate detailed command resource requirements, where: Represents resource demand, represents the resource sensitivity of the command prediction, is the adjustment coefficient; According to the detailed command resource requirements, the CPU and memory configuration of the smart speaker are adjusted using the formula: ; Generate an updated resource allocation, where: Indicates the updated resource allocation. is the adjustment parameter; Based on the updated resource allocation, monitor and record the running status after resource optimization, using the formula: ; Generates the speaker resource optimization state, where It is the speaker resource optimization state. is a performance evaluation parameter.

4. The smart speaker control system according to claim 3, It is characterized in that The steps for obtaining the optimized resource configuration are specifically as follows: According to the speaker resource optimization state, the formula is used: ; Generate the optimized state parameters after analysis, where represents the optimized state parameters after analysis, is the gain factor, is the baseline correction parameter; According to the optimized state parameters after the analysis, the required CPU and memory resources are calculated using the formula: ; Generates the evaluated resource parameters, where Represents the resource parameters after evaluation, , is the adjustment factor; Using the resource parameters evaluated above, the operating parameters are adjusted to optimize the sound control performance using the formula: ; Generate an optimized resource configuration, where: It is the optimized resource allocation. is the weight parameter, is to adjust the proportion, Represents the current resource configuration.

5. The smart speaker control system according to claim 4, It is characterized in that The steps for obtaining the audio state adjustment result are specifically as follows: Based on the optimized resource configuration, audio capture is activated to collect environmental audio data using the formula: ; Generate captured audio data parameters, where Represents the captured audio data, and are the gain and sensitivity adjustment factors, is the ambient audio intensity; The captured audio data parameters are input into the Kalman filter to analyze and update the state of the audio signal, adjust the smoothness and response speed of the signal, and use the formula: ; Generate a filtered audio signal where represents the filtered audio signal, and It is to adjust the sensitivity and threshold of signal processing; Based on the filtered audio signal, the audio state is updated using the formula: ; Generate an audio state adjustment result, where: is the result of adjusting the audio status. and Adjust the update rate and adjustment range of the audio status.

6. The smart speaker control system according to claim 5, It is characterized in that The steps for obtaining the optimized voice capture result are specifically as follows: According to the audio state adjustment result, real-time noise suppression is performed to optimize the audio capture in the environment, using the formula: ; Generate noise suppressed audio parameters, where is the optimized noise suppression parameter, They are noise suppression gain, stability coefficient and sensitivity adjustment parameters respectively; The audio parameters after noise suppression are used to capture and preliminarily process the voice commands controlled by the smart speaker, using the formula: ; Generate preliminary processed speech parameters, where is the voice command parameter after preliminary processing, and are the processing gain and weight parameters, The original voice data in the environment; According to the initially processed speech parameters, the speech recognition configuration of the smart speaker is updated using the formula: ; Generate optimized speech capture results, where For the optimized voice capture results, represents the update coefficient.

7. The smart speaker control system according to claim 6, It is characterized in that The steps for obtaining the speaker performance log are specifically as follows: Using the optimized voice capture results, the characterization of the response time is optimized, using the formula: ; Generate response time performance indicators, where and represents the gain factor, basic bias and adjustment factor, Represents the response time performance indicator; Combining the response time performance indicator and processing load, the formula is used: ; Generate performance log parameters, where is the performance log parameter, and is the weight parameter, Indicates processing load; Using the performance log parameters, by configuring the speaker parameters and recording key performance data, the formula is: ; Generates a speaker performance log, where is the fusion coefficient, This is the previous performance log. It is a log of speaker performance.

8. The smart speaker control system according to claim 7, It is characterized in that The steps for obtaining the adjusted speaker performance control response configuration are specifically as follows: Extract key performance indicators from the speaker performance log, including response efficiency and processing load, and calculate preliminary performance data using the formula: ; Generate performance analysis results, where Represents performance data extracted from logs, , Weight coefficient, Represents the processing load, Indicates response efficiency; Based on the performance analysis results, adjust the processing parameters of the speaker to optimize the response speed of the device, through the formula: ; Generate an optimized processing configuration where: For the optimized processing configuration, and is the adjustment parameter, Indicates adjusting the processing parameters of the speaker; Using the optimized processing configuration, update the speaker's performance control configuration and apply the formula: ; The adjusted speaker performance control response configuration is obtained, wherein: is the update coefficient, It is the previous speaker performance control configuration. Indicates the adjusted speaker performance control response configuration.