An AIUI background management system and method

Abstract

The present invention discloses an AIUI background management system and method, which specifically relates to the technical field of intelligent home-based elderly care services, and is used to solve the problem of poor home-based elderly care service experience; the present invention collects sound data through a high-sensitivity microphone system, uses spectrum subtraction and adaptive filtering technology to remove background noise and indoor echo, uses Mel-frequency cepstral coefficients to extract key acoustic features, and accurately classifies sound by combining a hybrid model of a convolutional neural network and a recurrent neural network to effectively capture local patterns of sound data, and then uses a hidden Markov model to perform behavioral pattern recognition on the classified sound data, dynamically adjusts energy use strategies, and then generates control impact information based on the results of behavioral pattern and energy use analysis, determines the actual situation of energy management through this information, and generates corresponding management signals, thereby improving the energy efficiency of the home environment and the comfort, convenience and safety of residents.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent home-based elderly care services, and more specifically, to an AIUI background management system and method. Background Art

[0002] With the increase in the number of elderly and disabled people, the traditional family-based elderly care model has problems such as mismatch between supply and demand of elderly care services, single service content, and weak home-based elderly care service team. It is urgent to break through the existing difficulties and optimize the home-based elderly care model by integrating emerging technologies;

[0003] Artificial intelligence (AI) technology simulates the human intelligence process through computer systems, including learning (obtaining information and rules through data and algorithms), reasoning (using rules to reach approximate or definite conclusions) and self-correction. In the field of home-based elderly care services, the application of artificial intelligence technology provides a variety of possibilities to enhance the quality and efficiency of home-based elderly care services. The interactive interface (UI) is used for human-computer interaction.

[0004] Deficiencies of existing technologies: Despite the introduction of intelligent monitoring and emergency response technologies based on artificial intelligence, existing home-based elderly care technologies have a low level of integration and insufficient intelligence. They are deficient in terms of response speed and accuracy in emergencies, and lack load management for equipment with the same working characteristics in daily situations. In addition, existing technology products lack a friendly user interface (UI) for elderly users, making it difficult for elderly users to operate independently, which in turn affects the intelligent and precise home-based elderly care service experience. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides an AIUI background management system and method to solve the problem of poor home-based elderly care service experience raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] An AIUI background management method includes the following steps:

[0008] Collect sound data, process background noise and indoor echo in the sound data, and use Mel frequency cepstrum coefficients to extract acoustic features of the processed sound data;

[0009] Combine the convolutional neural network with the recurrent neural network, and import the extracted acoustic features as data into the model for sound classification;

[0010] Use hidden Markov models to identify behavioral patterns in home environments based on sound classification data, manage energy based on behavioral patterns, and obtain control impact information generated during the energy management analysis of behavioral patterns;

[0011] Determine the energy management situation based on the control impact information and generate different signals, and adjust the energy management and behavior analysis strategies based on the generated signals.

[0012] In a preferred embodiment, sound data is collected and background noise and indoor echo in the sound data are processed. The specific process is as follows:

[0013] Set up sound receiving equipment in the area to collect sound data;

[0014] The collected sound data is subjected to noise filtering using a bandpass filter and adaptive noise suppression;

[0015] Spectral subtraction is used to remove background noise, and normalized minimum mean square error is used to process room echo.

[0016] In a preferred embodiment, the Mel frequency cepstrum coefficients are used to extract acoustic features from the processed sound data, and the specific process is as follows:

[0017] Perform high-pass filtering on the sound signal in the sound data, amplify the high-frequency part of the sound signal, and divide the continuous audio signal into short-time frames, and apply the Hamming window function to each short-time frame for windowing;

[0018] Perform fast Fourier transform on the windowed short time frame to convert the time domain signal into the frequency domain signal;

[0019] Apply the power spectrum of the frequency domain signal to a set of Mel filters and take the logarithm of the output of each Mel filter;

[0020] Discrete cosine transform is applied to the logarithmic Mel spectrum and the most important coefficients are retained as features.

[0021] In a preferred embodiment, a convolutional neural network is combined with a recurrent neural network, and the extracted acoustic features are fed into the model as data for sound classification. The specific process is as follows:

[0022] Organize and annotate a large-scale sound dataset, including household activity sounds and environmental background sounds;

[0023] Establish a classification model, input the extracted Mel frequency cepstral coefficients as input vectors into the selected model, and use supervised learning methods to train the model to recognize different types of sounds and behavior patterns;

[0024] Receive new sound input in real time, extract the features of Mel-frequency cepstral coefficients, input the extracted features into the classification model, and the classification model outputs the prediction of the sound category as sound classification data.

[0025] In a preferred embodiment, the control impact information generated during the energy management analysis of the behavior pattern is obtained, and the specific process is as follows:

[0026] Control impact information includes behavioral energy consumption synchronization information and response regulation information;

[0027] The behavior energy consumption synchronization information includes a multivariate behavior energy consumption synchronization index, and the response regulation information includes a dynamic behavior response energy regulation index;

[0028] The multivariable behavior energy consumption synchronization index and dynamic behavior response energy regulation index are obtained through joint calculation to obtain the management control coefficient.

[0029] In a preferred embodiment, the energy management situation is determined according to the control impact information and different signals are generated, and the energy management and behavior analysis strategy is adjusted according to the generated signals. The specific process is as follows:

[0030] Obtaining the management control coefficient obtained by the joint calculation, and comparing the management control coefficient with the synchronization threshold;

[0031] If the management control coefficient is greater than or equal to the synchronization threshold, a synchronization stability signal is generated to maintain the existing management strategy;

[0032] If the management control coefficient is less than the synchronization threshold, a linkage fluctuation signal is generated, and a warning or notification pops up through the UI interface to remind the user to pay attention to the problem and give suggestions for adjustment measures.

[0033] An AIUI background management system, used to implement the above-mentioned AIUI background management method, comprising:

[0034] The acquisition preprocessing module is used to collect sound data, process background noise and indoor echo in the sound data, and use Mel frequency cepstrum coefficients to extract acoustic features of the processed sound data;

[0035] The sound classification module is used to combine the convolutional neural network with the recurrent neural network and import the extracted acoustic features as data into the model for sound classification;

[0036] The behavioral energy joint module is used to identify the behavioral patterns of the home environment using the hidden Markov model for the sound classification data, manage the energy according to the behavioral patterns, and obtain the control impact information generated in the energy management analysis process of the behavioral patterns;

[0037] The management module is used to determine the energy management situation according to the control impact information and generate different signals, and adjust the energy management and behavior analysis strategies according to the generated signals.

[0038] Technical effects and advantages of the present invention:

[0039] The present invention collects sound data through a high-sensitivity microphone system, and effectively removes background noise and indoor echoes using bandpass filters, spectrum subtraction and adaptive filtering techniques, retaining useful acoustic signals. Then, the key acoustic features are extracted using Mel frequency cepstrum coefficients, and accurate sound classification is performed by combining a hybrid model of a convolutional neural network and a recurrent neural network, effectively capturing the local patterns and time series characteristics of the sound data, and then the classified sound data is subjected to behavioral pattern recognition using a hidden Markov model. The model can not only recognize conventional daily activities, but also detect abnormal behavior patterns in a timely manner, such as falling or emergency calls. According to the recognized behavior patterns, the energy use strategy can be dynamically adjusted to achieve optimal management of energy consumption. In addition, control impact information is generated based on the results of the behavior pattern and energy use analysis, and the actual situation of energy management is determined through this information, and corresponding management signals are generated accordingly. These signals further guide the adjustment of energy management and behavior monitoring strategies, thereby improving the energy efficiency of the home environment and the comfort, convenience and safety of residents. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 The present invention is a flowchart of an AIUI background management method.

[0041] Figure 2 The present invention is a schematic diagram of the structure of an AIUI background management system. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] Example 1: Figure 1 As shown, an AIUI background management method includes the following steps:

[0044] Collect sound data, process background noise and indoor echo in the sound data, and use Mel frequency cepstrum coefficients to extract acoustic features of the processed sound data;

[0045] Combine the convolutional neural network with the recurrent neural network, and import the extracted acoustic features as data into the model for sound classification;

[0046] Use hidden Markov models to identify behavioral patterns in home environments based on sound classification data, manage energy based on behavioral patterns, and obtain control impact information generated during the energy management analysis of behavioral patterns;

[0047] Determine the energy management situation based on the control impact information and generate different signals, and adjust the energy management and behavior analysis strategies based on the generated signals.

[0048] Monitor sounds and behaviors and integrate with intelligent robots to ensure effective recognition and response to sound activities in the user's home environment. The specific steps are as follows:

[0049] Conduct multi-device integration, i.e. design an integrated system to connect home service robots, indoor environment sensors, body sensor devices, mobile devices and cloud servers, and ensure that these devices can exchange data and respond to commands;

[0050] Configure the cloud server and establish it as a data processing and storage center to collect, analyze and share data from various devices, and update it to the remote caregiver's terminal in real time;

[0051] Collect and process sound data. Select high-sensitivity microphones and other sound receiving devices and place them in key locations in the home, such as the living room, kitchen, bedroom, etc. The microphones cover all important areas to capture sound without blind spots. At the same time, take into account the acoustic environment and avoid areas with overlapping sounds and echoes.

[0052] The collected sound data is processed. First, noise filtering is performed to reduce the impact of environmental noise and retain useful signals. Bandpass filters and adaptive noise suppression are used for noise filtering. ,in and are the low and high frequency cutoff frequencies of the filter respectively;

[0053] Use spectral subtraction to remove background noise. First, estimate the noise power spectral density, and then subtract the noise power spectrum from the power spectrum of the observed signal. The spectral subtraction formula is: ,in, is the power spectrum of the noisy audio signal, is the estimated noise power spectrum, It is a coefficient in spectral subtraction, which is used to adjust the strength of noise suppression;

[0054] Adaptive filtering technology is used to reduce indoor echo, that is, the normalized minimum mean square error (NLMS) algorithm is used. The specific formula is: ,in, are the current filter coefficients, is the input signal vector, is the error signal, μ is the step size, is a small positive number added for stability, is the energy of the input signal vector;

[0055] The sound data in the home environment is collected and processed to provide high-quality input data for subsequent sound recognition and behavior analysis. This refined processing helps to improve the overall performance of the subsequent energy management process and ensure the accuracy of sound activity monitoring and the timeliness of response.

[0056] After processing the collected sound data, sound recognition and behavior analysis are performed to monitor the user's behavior and health status through sound activities. The specific steps are as follows:

[0057] Acoustic feature extraction is performed, using Mel frequency cepstral coefficients (MFCC) as the main feature of the sound, and combining pitch, sound intensity, and long-term time domain features (such as Zero Crossing Rate) to enrich the feature set in order to better describe the content and emotional state of the sound signal, specifically:

[0058] Pre-emphasize and frame the sound signal in the sound data. Pre-emphasis is a high-pass filtering process used to amplify the high-frequency part of the signal. Pre-emphasis can help improve the spectrum distortion of the sound signal and make it more suitable for subsequent processing. The expression is: , where s(n) is the original signal, is the pre-emphasized signal, The value is usually between 0.95 and 0.97;

[0059] The continuous audio signal is divided into short time frames. Since the audio signal can be considered stable in a short time (such as 25ms), the frame calculation expression is: , where N is the number of samples per frame;

[0060] Apply the Hamming window function to each frame to reduce the signal discontinuity at the short-time frame boundary. The expression is: , where n ranges from 0 to N−1 and N is the window length;

[0061] Perform fast Fourier transform (FFT) on the windowed short time frame to convert the time domain signal into the frequency domain signal. , where k is the frequency index and N is the number of samples in each frame.

[0062] Apply the power spectrum of the FFT result to a set of Mel filters. The Mel filter is designed based on the different sensitivities of the human ear to sounds of different frequencies. The higher the frequency, the wider the bandwidth of the filter. The calculation expression is: ,in, is the output of the i-th Mel filter, is the gain of the i-th Mel filter at frequency k;

[0063] The logarithm of the output of each Mel filter is taken to simulate the nonlinear perception of human hearing, which increases the dynamic range and numerical stability of the signal;

[0064] Apply discrete cosine transform (DCT) to the logarithmic Mel spectrum to reduce the feature correlation between frames and retain the most important coefficients as features. The calculation expression is: , where is the jth MFCC coefficient, M is the number of Mel filters, and n is the filter index;

[0065] Through the above steps, useful features can be extracted from the original sound signal, which can be subsequently used for sound recognition and behavior analysis, thereby supporting intelligent monitoring of the elderly’s home environment.

[0066] The convolutional neural network (CNN) and the recurrent neural network (RNN) are combined to use the extracted acoustic features as data to establish a sound classification model. CNN is used to extract local patterns in sound features, while RNN (especially LSTM or GRU) is used to process the time series characteristics of sound data and capture time dependencies and long-distance dependencies. The specific steps are as follows:

[0067] Organize and annotate a large-scale sound dataset, including various household activity sounds and environmental background sounds. The data annotation should be classified in detail, such as conversations, TV sounds, kitchen activity sounds, emergency calls, etc.

[0068] Establish a classification model, use the extracted acoustic features (MFCC) as input vectors, input them into the selected model, and use supervised learning methods to train the model so that it can learn to recognize different types of sounds and behavior patterns. The simple expression of the hybrid CNN-RNN model is: ,in, is the input feature at time t, CNN represents the operation of convolutional neural network, LSTM represents the operation of long short-term memory network, is the hidden state at time t;

[0069] Real-time sound classification: receiving new sound input in real time, extracting its MFCC or other acoustic features, and inputting the extracted features into the trained classification model. The classification model will output the prediction of the sound category as the sound classification data, and perform corresponding actions or feedback according to the prediction results, such as triggering an alarm, starting corresponding home devices, or notifying family members or caregivers;

[0070] Use pattern recognition techniques (such as hidden Markov models or cluster analysis) to analyze long-term collected sound classification data, learn and identify common behavioral patterns in home environments, monitor the activities of users (such as the elderly), and promptly identify abnormal behaviors and real-time related monitoring equipment energy consumption, learn and predict behavioral patterns from time series data;

[0071] During data collection, sound classification data and / or sensor data such as motion are collected to record the state sequence of the elderly in their daily life;

[0072] Extract features from raw data, such as sound features (MFCC, etc.), motion intensity, activity type, etc.;

[0073] Perform state definition and define a set of hidden states, each of which represents a behavior mode, such as resting, walking, eating, falling, etc.

[0074] Establish a hidden Markov model, initialize the model parameters, and determine the state transition probability matrix A, expression: ,in, is the probability of transitioning from state i to state j, the observation probability matrix B, expression: ,in, is the probability of observing observation k in state j, with the initial state distribution C, expressed as: ,in, is the probability of being in state i at the initial moment;

[0075] Perform parameter estimation to maximize the likelihood probability of the observation sequence;

[0076] Use the Viterbi algorithm to find the most likely state sequence under a given observation sequence, i.e., the behavior pattern sequence. Based on the decoded state sequence, monitor and identify patterns in daily activities, such as normal daily activities and possible abnormal activities (such as falls).

[0077] Identify potential dangers or abnormal behaviors by analyzing abnormal patterns in state sequences (such as sudden state transitions). Analyzing behavior trends over time, such as increases or decreases in activity, can provide a basis for long-term health monitoring and preventive medicine. Once abnormal behavior is detected, an alarm can be automatically triggered to notify family members or medical service providers and initiate emergency response procedures.

[0078] Through these steps, the Hidden Markov Model can be effectively used to analyze and identify the behavioral patterns of the elderly, timely detect and respond to potential health risks and emergencies, and provide necessary help and intervention;

[0079] After analyzing the behavioral patterns, energy is managed based on the behavioral patterns, that is, behavioral pattern analysis is used to optimize home energy usage. In a home-based elderly care environment, the energy consumption of related equipment can be automatically adjusted according to the daily activities and behavioral habits of the residents.

[0080] Obtain control impact information generated during the energy management analysis of behavioral patterns. The control impact information helps analyze how devices are used under different behavioral patterns and determines the interaction between monitoring behaviors and device energy consumption. The control impact information includes behavioral energy consumption synchronization information and response adjustment information.

[0081] The behavior energy consumption synchronization information includes the multivariate behavior energy consumption synchronization index and is calibrated as MBESI, and the response regulation information includes the dynamic behavior response energy regulation index and is calibrated as DBREAI;

[0082] The multivariate behavioral energy consumption synchronization index in the behavioral energy consumption synchronization information is used to measure the synchronization and correlation between the behavioral patterns of the elderly and household energy consumption in a home-based elderly care environment. The multivariate behavioral energy consumption synchronization index quantifies the correlation between behavioral activities and energy use, thereby providing data support for energy management and behavioral monitoring;

[0083] The multivariate behavioral energy consumption synchronization index represents the synchronization between behavioral data and energy consumption data by comparing the time series of the two. By calculating the cosine similarity, time sliding average difference and activity peak synchronization between the behavioral frequency vector and the energy consumption vector, the multivariate behavioral energy consumption synchronization index reveals the statistical association and time alignment characteristics between behavioral patterns and energy consumption fluctuations;

[0084] Energy efficiency optimization: The multivariate behavioral energy consumption synchronization index can help home care service providers or family managers understand the energy usage patterns in the home environment and adjust energy consumption according to the behavior patterns of residents, such as automatically adjusting the use of lighting, heating or air conditioning to improve energy efficiency and reduce waste;

[0085] Behavior Monitoring and Safety: By monitoring the relationship between behavior and energy consumption, the Multivariate Behavioral Energy Consumption Synchronization Index helps identify possible abnormal or dangerous situations, such as high energy consumption occurring at unusual times that may indicate equipment failure or unexpected events.

[0086] The multivariate behavior energy consumption synchronization index is obtained as follows:

[0087] Get the behavior frequency and energy consumption value within the set time window k, convert the behavior frequency into the behavior data vector PA, convert the energy consumption value into the energy consumption data vector PB, calculate the dot product value PAB, and the calculation expression is: , where , are the elements of vectors PA and PB at the i-th position, n is the number of elements, and the norm of the behavior frequency vector is calculated: , calculate the energy consumption vector norm: ;

[0088] For each point in the time window k, calculate the moving average of the behavior frequency in the time window: And the energy consumption moving average: , where t is the current time point, and the sliding difference value is calculated: , where N is the number of total time windows. The multivariate behavior energy consumption synchronization index is calculated using the following expression: .

[0089] It should be noted that the time window is determined according to different actual needs and scenarios, such as hourly, daily or weekly, and the choice of window depends on the type of behavior and the purpose of energy consumption monitoring. For example, daytime activities may be analyzed by hour, while nighttime sleep patterns may be analyzed by day; behavioral data can be collected from various sensors deployed in the living environment, such as motion sensors, door and window sensors, mattress sensors, etc. These sensors can record the frequency and type of activities of the elderly, such as walking, sitting, standing up, entering and leaving the room, etc. For each time window, the number of occurrences of a specific behavior is counted, and the number of behaviors is divided by the length of the time window (hours or days) to obtain the average behavior frequency of the window; energy consumption data is usually provided by smart meters or various energy monitoring devices, which can record the power consumption of various household appliances (such as lights, air conditioners, TVs, refrigerators, etc.), and for each time window, the energy consumption of all related appliances is accumulated.

[0090] The dynamic behavior response energy regulation index in the response regulation information represents the time alignment and adjustment accuracy between the mode change and the energy system adjustment, which is used to determine how energy management can respond to the user's behavior changes in a timely and effective manner.

[0091] The Dynamic Behavior Response Energy Regulation Index evaluates the time required from a significant change in behavior pattern to the energy system's response (such as adjusting temperature, lighting, etc.); a fast response time means that the system can quickly adapt to the needs of users, increasing the comfort and satisfaction of residents; not only a fast response, the Dynamic Behavior Response Energy Regulation Index also evaluates the accuracy of energy system adjustments. For example, if users are less active at night, does the system appropriately reduce the use of lighting and heating? This is related to the efficient use of energy and energy saving.

[0092] The dynamic behavior response energy regulation index helps reduce unnecessary energy waste by ensuring that energy consumption closely corresponds to the actual behavior and needs of users. For example, it automatically turns off the lights and other appliances in a room when the user is not in the room; by sensitively responding to user behavior, it can ensure that the living environment (such as temperature, humidity and light) is always kept in the most comfortable state. This is particularly important for improving the quality of home-based elderly care;

[0093] In short, the dynamic behavior response energy regulation index is a key indicator that not only improves the intelligence and automation level of the home-based elderly care environment, but also optimizes energy use, improves the quality of life of residents, and has a positive impact on the economy and the environment. By continuously monitoring and optimizing dynamic behavior response energy regulation, home-based elderly care equipment can achieve higher service levels and operational efficiency.

[0094] The dynamic behavior response energy regulation index is obtained as follows:

[0095] Perform differential processing on the real-time collected behavior frequency and energy consumption value to obtain the behavior differential value: Difference in energy consumption: ,in, represents the frequency of behavior at time T, Represents the energy consumption value at time T; set the moving window size D, calculate the seasonal moving average component, and the calculation expression is: , where L represents the number of data points on both sides of the window center point;

[0096] Get the average of the behavioral difference values Average value of energy consumption difference , calculate the cross-correlation value, the calculation expression is: , calculate the dynamic behavior response energy regulation index, the calculation expression is: .

[0097] It should be noted that first-order differencing is usually sufficient to remove most linear trends. If there are more complex nonlinear trends in the data, second-order or higher-order differencing may be required; the moving window size should be chosen to match the length of the main seasonal cycle of the data. For example, if the data shows annual seasonal changes, the window size may be set to 12 (monthly data).

[0098] The multivariable behavior energy consumption synchronization index and dynamic behavior response energy regulation index are obtained by joint calculation to obtain the management control coefficient, which is expressed as follows: , where is the management control coefficient, , is the preset proportional coefficient of the multivariable behavior energy consumption synchronization index MBESI and the dynamic behavior response energy regulation index DBREAI, and , Both are greater than 0.

[0099] It should be noted that the size of the preset proportional coefficient is a specific value obtained by quantifying each parameter. In order to facilitate subsequent comparison, the size of the coefficient depends on the amount of sample data and the preliminary setting of the corresponding preset proportional coefficient for each set of sample data by technical personnel in this field. It is not unique, as long as it does not affect the proportional relationship between the parameter and the quantified value, just like the dynamic behavior response energy regulation index is proportional to the management control coefficient.

[0100] The larger the multivariate behavior energy consumption synchronization index and the dynamic behavior response energy regulation index, that is, the larger the performance value of the management regulation coefficient, indicates that the behavior monitoring and energy consumption management in the home-based elderly care environment are very effective in adapting to and supporting the actual needs of users, which means that collaborative work can be carried out to achieve refined management, indicating that energy use has been effectively optimized and energy consumption can be dynamically adjusted according to the user's behavior pattern, thereby reducing waste; efficient operation of equipment management can not only save energy, but also enhance user comfort and satisfaction, because environmental factors (such as temperature and light) are adjusted according to the actual activities and needs of users. Improving energy efficiency and optimizing energy management usually bring economic benefits, such as reducing energy costs, and at the same time, the negative impact on the environment is also smaller;

[0101] The smaller the multivariate behavior energy consumption synchronization index is, the smaller the dynamic behavior response energy regulation index is, that is, the smaller the performance value of the management regulation coefficient is. The smaller the multivariate behavior energy consumption synchronization index is, the less synchronization there is between the behavior data and energy consumption. This may mean that the energy consumption management system cannot accurately adjust energy consumption according to the actual activities of the user, which may lead to energy waste or unnecessary energy consumption.

[0102] A small dynamic behavior response energy regulation index indicates that the device is slow to respond to behavior changes when monitoring, and cannot adjust the environmental settings in time to meet user needs, which may affect user comfort and satisfaction;

[0103] Because energy consumption is not synchronized with behavioral activities and the response is slow, energy may be consumed when it is not needed, such as continuously running air conditioning or lighting in unoccupied rooms; lower energy efficiency and high energy consumption will increase the operating costs of the home care environment, including higher energy bills and possible maintenance costs.

[0104] Compare the generated management control coefficient with the synchronization threshold to generate different management signals, and use different strategies to adjust the home-based elderly care monitoring equipment and behavior analysis according to the generated management signals;

[0105] After obtaining the management control coefficient, compare the management control coefficient with the synchronization threshold;

[0106] If the management control coefficient is greater than or equal to the synchronization threshold, a synchronization stability signal is generated, which shows that the collaboration between behavior monitoring and energy consumption management is very effective, indicating that the linkage between related behaviors and energy consumption can accurately adjust energy consumption according to the user's behavior pattern, such as automatically adjusting lighting and temperature control to meet the actual needs of users. This also reflects the high accuracy of the behavior prediction model, which can timely and accurately identify the user's activity pattern, thereby prompting corresponding adjustments in energy management;

[0107] If the management control coefficient is less than the synchronization threshold, a linkage fluctuation signal is generated, indicating that the collaborative work between behavior monitoring and energy management is not effective. This may mean that energy management cannot be adjusted accurately or timely according to the user's behavior pattern, the response to user behavior changes is slow or the response measures are not accurate enough, resulting in the inability to effectively adjust energy consumption to meet user needs;

[0108] Maintain current monitoring and energy management settings while generating synchronous stabilization signals, and perform continuous monitoring and periodic evaluation to maintain system performance and respond to possible behavioral changes or equipment aging issues;

[0109] For the generation of linkage fluctuation signals, a comprehensive review of existing behavior monitoring and energy consumption management is conducted to identify performance bottlenecks or deficiencies and make necessary adjustments or upgrades. It may be necessary to improve data analysis algorithms to more accurately predict energy consumption requirements from behavior data.

[0110] Design dedicated areas or icons in the UI interface to display the generated signals. For example, use different colored light icons to represent different management signals, and provide click or hover functions to view detailed device values ​​and explanations;

[0111] When a "linked fluctuation signal" is generated, a warning or notification pops up through the UI interface to remind users of possible problems and recommended adjustment measures. For a "synchronous stable signal", a confirmation message or prompt can be provided to inform users that the relevant equipment is operating normally;

[0112] Provide feedback options in the UI interface to allow users to report problems and make suggestions. Based on different management signals, provide specific action suggestions or automatic adjustment solutions, such as adjusting temperature control settings or checking specific equipment.

[0113] It should be noted that the synchronization threshold can be set according to specific scenarios and needs, and is usually adjusted and optimized based on factors such as historical data and real-time data.

[0114] The present invention collects sound data through a high-sensitivity microphone system, and effectively removes background noise and indoor echoes using bandpass filters, spectrum subtraction and adaptive filtering techniques, retaining useful acoustic signals. Then, the key acoustic features are extracted using Mel frequency cepstrum coefficients, and accurate sound classification is performed by combining a hybrid model of a convolutional neural network and a recurrent neural network, effectively capturing the local patterns and time series characteristics of the sound data, and then the classified sound data is subjected to behavioral pattern recognition using a hidden Markov model. The model can not only recognize conventional daily activities, but also detect abnormal behavior patterns in a timely manner, such as falling or emergency calls. According to the recognized behavior patterns, the energy use strategy can be dynamically adjusted to achieve optimal management of energy consumption. In addition, control impact information is generated based on the results of the behavior pattern and energy use analysis, and the actual situation of energy management is determined through this information, and corresponding management signals are generated accordingly. These signals further guide the adjustment of energy management and behavior monitoring strategies, thereby improving the energy efficiency of the home environment and the comfort, convenience and safety of residents.

[0115] Embodiment 2: This embodiment is a system embodiment of embodiment 1, which is used to implement an AIUI background management method introduced in embodiment 1, such as Figure 2 As shown, specifically including:

[0116] An AIUI background management system, used to implement the above-mentioned AIUI background management method, comprising:

[0117] The acquisition preprocessing module is used to collect sound data, process background noise and indoor echo in the sound data, and use Mel frequency cepstrum coefficients to extract acoustic features of the processed sound data;

[0118] The sound classification module is used to combine the convolutional neural network with the recurrent neural network and import the extracted acoustic features as data into the model for sound classification;

[0119] The behavioral energy joint module is used to identify the behavioral patterns of the home environment using the hidden Markov model for the sound classification data, manage the energy according to the behavioral patterns, and obtain the control impact information generated in the energy management analysis process of the behavioral patterns;

[0120] The management module determines the energy management situation and generates different signals based on the control impact information, and adjusts the energy management and behavior analysis strategies based on the generated signals.

[0121] The above formulas are all dimensionless and calculated numerically. Specific dimension removal can be achieved by various means such as standardization, which will not be elaborated here. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0122] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, an ATA hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium may be a solid-state ATA hard disk.

[0123] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0124] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0125] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0126] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0127] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0128] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An AIUI background management method, characterized in that: The steps include: Collect sound data, process background noise and indoor echo in the sound data, and use Mel frequency cepstrum coefficients to extract acoustic features of the processed sound data; Combine the convolutional neural network with the recurrent neural network, and import the extracted acoustic features as data into the model for sound classification; Use hidden Markov models to identify behavioral patterns in home environments based on sound classification data, manage energy based on behavioral patterns, and obtain control impact information generated during the energy management analysis of behavioral patterns; Determine the energy management situation and generate different signals based on the control impact information, and adjust the energy management and behavior analysis strategies based on the generated signals; Obtain the control impact information generated during the energy management analysis of the behavioral model. The specific process is as follows: Control impact information includes behavioral energy consumption synchronization information and response regulation information; The behavior energy consumption synchronization information includes a multivariate behavior energy consumption synchronization index, and the response regulation information includes a dynamic behavior response energy regulation index; The multivariate behavioral energy consumption synchronization index represents the correlation between behavioral activities and energy use; The dynamic behavior response energy regulation index represents the time alignment and adjustment accuracy between the mode change and the energy system adjustment; The multivariate behavior energy consumption synchronization index is obtained as follows: Get the behavior frequency and energy consumption value within the set time window k, convert the behavior frequency into the behavior data vector PA, convert the energy consumption value into the energy consumption data vector PB, calculate the dot product value PAB, and the calculation expression is: , where , are the elements of vectors PA and PB at the i-th position, n is the number of elements, and the norm of the behavior frequency vector is calculated: , calculate the energy consumption vector norm: ; For each point in the time window k, calculate the moving average of the behavior frequency in the time window: And the energy consumption moving average: , where t is the current time point, and the sliding difference value is calculated: , where N is the number of total time windows. The multivariate behavior energy consumption synchronization index is calculated using the following expression: ; The dynamic behavior response energy regulation index is obtained as follows: Perform differential processing on the real-time collected behavior frequency and energy consumption value to obtain the behavior differential value: Difference in energy consumption: ,in, represents the frequency of behavior at time T, Represents the energy consumption value at time T; set the moving window size D, calculate the seasonal moving average component, and the calculation expression is: , where L represents the number of data points on both sides of the window center point; Get the average of the behavioral difference values Average value of energy consumption difference , calculate the cross-correlation value, the calculation expression is: , calculate the dynamic behavior response energy regulation index, the calculation expression is: ; The multivariable behavior energy consumption synchronization index and dynamic behavior response energy regulation index are obtained through joint calculation to obtain the management control coefficient.

2. The AIUI background management method according to claim 1, characterized in that: Collect sound data and process the background noise and indoor echo in the sound data. The specific process is as follows: Set up sound receiving equipment in the area to collect sound data; The collected sound data is subjected to noise filtering using a bandpass filter and adaptive noise suppression; Spectral subtraction is used to remove background noise, and normalized minimum mean square error is used to process room echo.

3. The AIUI background management method according to claim 2, characterized in that: The Mel frequency cepstrum coefficients are used to extract acoustic features from the processed sound data. The specific process is as follows: Perform high-pass filtering on the sound signal in the sound data, amplify the high-frequency part of the sound signal, and divide the continuous audio signal into short-time frames, and apply the Hamming window function to each short-time frame for windowing; Perform fast Fourier transform on the windowed short time frame to convert the time domain signal into the frequency domain signal; Apply the power spectrum of the frequency domain signal to a set of Mel filters and take the logarithm of the output of each Mel filter; Discrete cosine transform is applied to the logarithmic Mel spectrum and the most important coefficients are retained as features.

4. The AIUI background management method according to claim 3, characterized in that: The convolutional neural network is combined with the recurrent neural network, and the extracted acoustic features are imported into the model as data for sound classification. The specific process is as follows: Organize and annotate a large-scale sound dataset, including household activity sounds and environmental background sounds; Establish a classification model, input the extracted Mel frequency cepstral coefficients as input vectors into the selected model, and use supervised learning methods to train the model to recognize different types of sounds and behavior patterns; Receive new sound input in real time, extract the features of Mel-frequency cepstral coefficients, input the extracted features into the classification model, and the classification model outputs the prediction of the sound category as sound classification data.

5. The AIUI background management method according to claim 4, characterized in that: The energy management situation is determined based on the control impact information and different signals are generated. The energy management and behavior analysis strategies are adjusted based on the generated signals. The specific process is as follows: Obtaining the management control coefficient obtained by the joint calculation, and comparing the management control coefficient with the synchronization threshold; If the management control coefficient is greater than or equal to the synchronization threshold, a synchronization stability signal is generated to maintain the existing management strategy; If the management control coefficient is less than the synchronization threshold, a linkage fluctuation signal is generated, and a warning or notification pops up through the UI interface to remind users of possible problems and recommended adjustment measures.

6. An AIUI background management system, used to implement an AIUI background management method according to any one of claims 1 to 5, characterized in that: include: The acquisition preprocessing module is used to collect sound data, process background noise and indoor echo in the sound data, and use Mel frequency cepstrum coefficients to extract acoustic features of the processed sound data; The sound classification module is used to combine the convolutional neural network with the recurrent neural network and import the extracted acoustic features as data into the model for sound classification; The behavioral energy joint module is used to identify the behavioral patterns of the home environment using the hidden Markov model for the sound classification data, manage the energy according to the behavioral patterns, and obtain the control impact information generated in the energy management analysis process of the behavioral patterns; The management module is used to obtain the control influence information generated in the process of matching the digital personality response with the user's current situation, determine the matching situation and generate a signal based on the control influence information, and perform corresponding mode strategy management based on the generated signal.

Images