A control method and system for acoustic early education machine

Abstract

The present invention discloses a control method and system for an acoustic early childhood education machine, which specifically relates to the field of intelligent speech recognition, and is used to solve the problem of adaptive adjustment of speech recognition sensitivity in complex environments. It combines multi-microphone array technology with directional filtering and spectrum segmentation to accurately capture the sound source environment and generate a sound source characteristic vector containing information such as distance, direction, and frequency, providing an accurate basis for noise evaluation and sensitivity adjustment. Through feature clustering and time-frequency domain energy analysis, combined with adaptive dynamic mapping of noise change rate, noise fluctuations and change trends are evaluated in real time, and a noise reference level adapted to different environments is generated to ensure the accuracy and stability of speech recognition. By combining the noise reference level with the change rate, the duration of the speech detection window and the pickup range are adaptively optimized to improve the responsiveness and environmental adaptability of speech recognition, ensure that the device works efficiently in a complex home environment, and provide stable and accurate voice control.

Description

Technical Field

[0001] The present invention relates to the field of intelligent speech recognition, and more specifically, to a control method and system for an acoustic early childhood education machine. Background Art

[0002] In modern family education, acoustic early education machines, as a smart device that helps children's early learning and cognitive development, have gradually become the first choice of many parents. Acoustic early education machines use voice recognition and command feedback functions to allow children to independently select interactive educational resources such as stories, nursery rhymes, and cognitive content through simple voice commands. This way of autonomous learning through voice control can not only satisfy children's curiosity, but also cultivate their early language expression and cognitive abilities. However, the complexity of the family environment, such as the conversations of family members, background sound sources (such as TV and home appliance sounds), and external environmental noise, all pose severe challenges to the voice recognition performance of acoustic early education machines. Existing technologies usually set a constant sensitivity threshold to keep the device working normally in most situations, but it is difficult to accurately adjust the sensitivity in different environments, affecting the recognition accuracy and battery life of the device.

[0003] Current acoustic early childhood education machines generally lack adaptive adjustment for different sound source distances, volume levels, and usage scenarios. The device usually operates continuously in a relatively high sensitivity mode to ensure that children's commands can be recognized even at a long distance or at a low volume. However, in a changing home environment, this constant high sensitivity design can easily lead to frequent false triggering of the device, and even without children's commands, it may be repeatedly activated due to environmental noise. This lack of a scenario-based sensitivity dynamic adjustment mechanism causes the device to misidentify in complex environments, resulting in slow response and reduced recognition accuracy, which in turn affects the child's interactive experience with the device.

[0004] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiment of the present invention provides a control method and system for an acoustic early education machine, which realizes accurate capture of the sound source environment by combining multi-microphone array technology with directional filtering and spectrum segmentation, and generates a sound source characteristic vector containing the distance, direction and frequency information of the sound source, thereby providing an accurate basis for subsequent environmental noise evaluation and sensitivity adjustment. By combining feature clustering with time-frequency domain energy analysis, adaptive dynamic mapping based on the noise change rate can evaluate the noise volatility and change trend of the current environment in real time, and generate a noise reference level adapted to different typical scenes. It ensures that the device can accurately recognize voice commands in a variable noise environment while reducing the interference of external noise. In addition, by combining the noise reference level and the change rate, an adaptive window optimization and power adaptation algorithm are adopted to dynamically adjust the voice detection window duration and pickup range, thereby maintaining the accuracy of voice recognition while ensuring the stability and responsiveness of the device in a complex home environment. Overall, it can provide efficient and stable voice control under a complex noise background, greatly improve the user experience and intelligence level of the device, and solve the problem that the sensitivity of traditional technology cannot be adaptively adjusted in a variable environment, so as to solve the problems raised in the above background technology.

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

[0007] A control method for an acoustic early childhood education machine comprises the following steps:

[0008] S1, based on a multi-microphone array and combined with directional filtering and spectrum segmentation technology, obtains the multi-dimensional audio characteristics of the sound source environment and generates a sound source characteristic vector containing the sound source distance, direction, and frequency information.

[0009] S2, based on the sound source characteristic vector, through feature clustering and time-frequency domain energy analysis, combined with adaptive dynamic mapping of the noise change rate, evaluates the volatility and change trend of the current environmental noise, and generates a noise benchmark level suitable for different typical scenarios.

[0010] S3 dynamically adjusts the speech detection window duration and pickup range by combining the noise baseline level and change rate.

[0011] In a preferred embodiment, step S1 includes the following contents:

[0012] S1.1, use multiple built-in microphones of the device to form an array. Assume that the spatial position of each microphone is known. Each microphone records the arrival time of sound waves in different directions to form an initial sound collection matrix based on spatial distribution. The rows of the matrix represent microphone channels, and the columns represent audio signal values ​​that change over time.

[0013] S1.2, based on the time difference of microphone signal arrival, a spatial sound source localization model is constructed, and the cross-correlation between signals is calculated using the generalized cross-power spectrum method to obtain the time difference matrix of each microphone pair.

[0014] S1.3, based on the signal after directional filtering, the attenuation model is used to estimate the average distance between the sound source and the microphone array.

[0015] S1.4, perform Fourier transform on the signal after directional filtering to obtain spectrum information; use spectrum segmentation method to decompose the signal into multiple frequency bands, and screen out the main frequency peaks to obtain frequency energy concentration; through the multiple frequency bands segmented, extract the main frequency peaks and frequency band energy proportions of the sound source to form characteristic frequencies.

[0016] S1.5, comprehensively analyzing the processing results, combining the average distance, directional angle and pitch angle, characteristic frequency, and frequency concentration parameters of the sound source into a sound source characteristic vector.

[0017] In a preferred embodiment, step S2 includes the following contents:

[0018] S2.1, first, cluster the sound source characteristic vectors to obtain the cluster category to which each sound source belongs; on this basis, the adaptive mapping model of time-frequency domain distribution is used to perform deep mapping on the clustering results, establish the multidimensional feature space of each noise source, and set each noise source category of the clustering results as , and its mapping matrix in the feature space is ; The generation formula of the mapping matrix is: ;in, For the The audio feature vector of cluster samples, are the corresponding time and frequency dimension features, represents the tensor product of a vector and a matrix, For the The number of class samples.

[0019] S2.2, after feature clustering, extract the time domain and frequency domain features of each noise source category; evaluate the volatility and distribution trend of noise through the instantaneous energy change in the time domain and the spectrum drift characteristics in the frequency domain; set the time domain energy and frequency domain energy The calculation formula is as follows: ; ;in, is the time signal at time The amplitude of The spectrum signal at frequency The amplitude of and Represents the number of sampling points in time and frequency respectively, Represents the sampling point index of the time domain signal, which is used for time domain energy calculation. Represents the frequency point index of the frequency domain signal, which is used for frequency domain energy calculation; by combining time domain energy and frequency domain energy, a multidimensional energy function is constructed , and dynamically track the energy of the noise source: ;in, and is the dynamically adaptive scaling factor.

[0020] In a preferred embodiment, step S2 further includes the following contents:

[0021] S2.3, after extracting the time-frequency domain features, calculate the volatility and change rate of each noise source category; through the incremental analysis of the time-frequency features, define the noise change rate as the incremental ratio of the noise energy, and the specific formula is: ;in, Indicates the rate of change of noise energy.

[0022] S2.4, based on the clustering results and the time-frequency domain energy distribution and change rate analysis of the noise source, the characteristics of each noise source are compared with the noise baseline level. Correspondingly, the noise floor level is calculated by the following generation logic: ;in, is the number of cluster categories, is the adaptive dynamic mapping function, and Represent the weighting functions of the clustering matrix and the noise change rate respectively.

[0023] In a preferred embodiment, step S3 includes the following contents:

[0024] S3.1, adjust the window length The adjustment is based on the interaction between the noise baseline level and the noise change rate, reflecting the complexity and change trend of environmental noise. The specific calculation process is as follows: ;in, Indicates the noise baseline level at the current moment; Indicates the noise change rate at the current moment; is a small constant to prevent the window length from being zero.

[0025] In a preferred embodiment, step S3 further includes the following contents:

[0026] S3.2, pickup range The adjustment is based on the interaction between the noise floor level and the noise change rate, aiming to optimize the sensitivity and coverage of the microphone array to ensure the accuracy of sound source identification and response efficiency. The specific calculation process is as follows: ;in, is a constant used to smooth the effects of the rate of change of noise.

[0027] A control system for an acoustic early childhood education machine comprises a sound source detection module, a noise assessment module and a sensitivity control module.

[0028] Sound source detection module: Based on a multi-microphone array, combined with directional filtering and spectrum segmentation technology, it obtains multi-dimensional audio features in the environment and generates a sound source characteristic vector containing sound source distance, direction and frequency information; the sound source characteristic vector is passed to the subsequent noise assessment module.

[0029] Noise assessment module: Through the sound source characteristic vector, feature clustering and time-frequency domain energy analysis are adopted, combined with the adaptive dynamic mapping of the noise change rate, to evaluate the noise volatility and change trend of the current environment, and generate the noise reference level suitable for different typical scenarios; the noise reference level is passed to the sensitivity control module.

[0030] Sensitivity control module: dynamically adjusts the voice detection window duration and pickup range based on the noise baseline level and change rate.

[0031] The technical effects and advantages of the control method and system of an acoustic early childhood education machine of the present invention are as follows:

[0032] The present invention combines multi-microphone array technology with directional filtering and spectrum segmentation to accurately capture the sound source environment and generate a sound source characteristic vector containing the sound source distance, direction, and frequency information, thereby providing an accurate basis for subsequent environmental noise evaluation and sensitivity adjustment. By combining feature clustering with time-frequency domain energy analysis, adaptive dynamic mapping based on the noise change rate can evaluate the noise volatility and change trend of the current environment in real time, and generate a noise reference level adapted to different typical scenes. This processing process ensures that the device can accurately identify voice commands in a variable noise environment while reducing the interference of external noise. And by combining the noise reference level and the change rate, the adaptive window optimization and power adaptation algorithm are used to dynamically adjust the voice detection window duration and pickup range, thereby maintaining the accuracy of voice recognition while ensuring the stability and responsiveness of the device in a complex home environment. Overall, the device can improve the responsiveness and environmental adaptability of voice recognition under a complex noise background, ensure that the device works efficiently in a complex home environment, and provide stable and accurate voice control. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1The present invention is a flowchart of a control method for an acoustic early childhood education machine.

[0034] Figure 2 It is a structural schematic diagram of a control system of an acoustic early education machine of the present invention. DETAILED DESCRIPTION

[0035] 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.

[0036] Embodiment 1: Figure 1 The present invention provides a control method for an acoustic early childhood education machine, comprising:

[0037] S1, based on a multi-microphone array and combined with directional filtering and spectrum segmentation technology, obtains the multi-dimensional audio characteristics of the sound source environment and generates a sound source characteristic vector containing the sound source distance, direction, and frequency information.

[0038] S2, based on the sound source characteristic vector, through feature clustering and time-frequency domain energy analysis, combined with adaptive dynamic mapping of the noise change rate, evaluates the volatility and change trend of the current environmental noise, and generates a noise benchmark level suitable for different typical scenarios.

[0039] S3 dynamically adjusts the speech detection window duration and pickup range by combining the noise baseline level and change rate.

[0040] In the ever-changing home application environment of acoustic early education machines, the device must accurately identify and distinguish sound sources from different directions, distances and volumes in order to effectively capture children's voice commands while eliminating the interference of background noise. To achieve this, a multi-microphone array is used to obtain the multi-dimensional audio features of the sound source through directional filtering and spectrum segmentation technology, and generate a sound source characteristic vector containing the distance, direction and frequency information of the sound source. This processing process can not only accurately identify the characteristics of the sound source, but also provide high-quality input data for subsequent noise baseline determination.

[0041] Step S1 includes the following contents:

[0042] S1.1, using multiple microphones built into the device to form an array, assuming that the spatial position of each microphone is known. Each microphone records the arrival time of sound waves from different directions to form an initial sound collection matrix based on spatial distribution , the rows of the matrix represent microphone channels and the columns represent the audio signal values ​​changing over time. This acquisition matrix is ​​pre-processed to filter out obvious electrical noise and sorted by the time difference received by the microphone.

[0043] S1.2, based on the time difference of microphone signal arrival, a spatial sound source localization model is constructed, and the cross-correlation between signals is calculated using the generalized cross power spectrum method (GCC-PHAT) to obtain the time difference matrix of each microphone pair : ;in , represents the channel pair of the microphone, is the time delay. The azimuth and elevation angles of the sound source are inversely solved through the time difference matrix , preliminarily determine the direction of the sound source.

[0044] The directional filter forms an angle screening window based on the angle information, retaining only the main sound source signal in a specific direction and reducing the interference of excess noise.

[0045] S1.3, based on the signal after directional filtering, use the attenuation model to estimate the average distance between the sound source and the microphone array. Assuming that the volume of the sound source is constant, the signal strength received by each microphone is and distance Inversely proportional, calculate the average distance: ;in, The average value of the distances measured by multiple microphones is taken to obtain the estimated distance of the sound source, which is used to construct the distance parameter in the sound source characteristic vector. is the number of microphones in the array used to estimate the distance to the sound source.

[0046] S1.4, perform Fourier transform on the signal after directional filtering to obtain spectrum information. Use spectrum segmentation method to decompose the signal into multiple frequency bands, filter out the main frequency peaks (characteristic frequencies), and calculate the frequency energy concentration. Frequency concentration Defined as the proportion of main frequency energy in the spectrum, it is used to characterize the stability of the sound source: ;in is the characteristic frequency, is the frequency bandwidth, is the spectrum signal.

[0047] Through the multiple frequency bands divided out, the main frequency peaks and frequency band energy ratios of the sound source are further extracted to form spectral features, which are used to characterize the audio characteristics of the sound source.

[0048] S1.5, combining the processing results, the average distance of the sound source , azimuth and elevation , characteristic frequency , frequency concentration The same parameters are combined into the sound source characteristic vector: ; The sound source characteristic vector includes the spatial position, audio characteristics and spectrum distribution of the sound source, providing an accurate description of the sound field for subsequent processing steps.

[0049] Step S1 accurately extracts and analyzes the multi-dimensional audio features of the sound source through the construction of a multi-microphone array, combined with directional filtering, time difference calculation and spectrum segmentation technology, and forms a sound source characteristic vector containing information such as the distance, direction, and frequency of the sound source. This characteristic vector combines the spatial position and audio features, can effectively distinguish between children's voice commands and environmental noise, and realize accurate identification and dynamic tracking of sound sources in complex family scenes. Through multi-level feature extraction, the acoustic early childhood education machine can adapt to the changing environmental sound field, thereby providing accurate input data for the subsequent determination of the noise benchmark, and laying a reliable technical foundation for improving speech recognition accuracy and equipment endurance. This ensures the adaptive ability of the acoustic early childhood education machine in complex environments, and significantly enhances the practicality and intelligence of the equipment in practical applications.

[0050] In the use scenarios of acoustic early childhood education machines, the device needs to cope with a variety of noise environments to achieve accurate voice recognition and response. Children's voice commands are often issued in complex background noises, and the noise sources may include the conversations of family members, the background sound of the TV or stereo, and even the noise of external traffic. The presence of these noises not only affects the accuracy of the device's voice recognition, but may also cause misoperation or delayed device response. Therefore, how to dynamically adjust the device's voice control sensitivity in a noisy environment to avoid misrecognition of voice and ensure the stability of the device has become a key issue in improving user experience.

[0051] To meet this challenge, this step uses innovative sound source characteristic vector analysis technology, combined with multi-dimensional feature clustering and time-frequency domain energy analysis, to intelligently identify and quantify the noise characteristics in the current environment. By real-time evaluation of the volatility and change rate of the noise environment, a noise baseline level that can adapt to different typical scenarios is generated to ensure that the device can make adaptive adjustments under various noise conditions, accurately grasp voice commands, and improve the responsiveness and stability of voice control.

[0052] Step S2 includes the following contents:

[0053] S2.1, first, cluster the sound source characteristic vectors (for example, through K-means or density peak clustering) to obtain the cluster category to which each sound source belongs. On this basis, the adaptive mapping model of time-frequency domain distribution is used to perform deep mapping on the clustering results and establish the multi-dimensional feature space of each noise source (including time, frequency, directionality and other dimensions). Set each noise source category of the clustering result to (in Indicates type noise source), its mapping matrix in the feature space is .

[0054] The generation formula of the mapping matrix is: ;in, For the The audio feature vector of cluster samples, are the corresponding time and frequency dimension features, represents the tensor product of a vector and a matrix, For the The number of class samples.

[0055] S2.2, after feature clustering, extract the complex features of each noise source category in the time domain and frequency domain. Evaluate the volatility and distribution trend of noise through the instantaneous energy change in the time domain and the spectrum drift characteristics in the frequency domain.

[0056] Setting time domain energy and frequency domain energy The calculation formula is as follows: ; ;in, is the time signal at time The amplitude of The spectrum signal at frequency The amplitude of and Represents the number of sampling points in time and frequency respectively, Represents the sampling point index of the time domain signal, which is used for time domain energy calculation. Represents the frequency point index of the frequency domain signal, which is used for frequency domain energy calculation. By combining time domain energy and frequency domain energy, a multidimensional energy function is constructed. , and dynamically track the energy of the noise source: ;in, and is a dynamically adaptive scaling factor that adjusts the feature contribution in the time-frequency domain by reflecting the nonlinear characteristics of the signal strength.

[0057] S2.3, after extracting the time-frequency domain features, further calculate the volatility and change rate of each noise source category. Through the incremental analysis of the time-frequency features, the noise change rate is defined as the incremental ratio of the noise energy. The specific formula is: ; In this formula, Indicates the rate of change of noise energy. By associating the fluctuation and increment ratio in each time period, the noise fluctuation rate is defined as a key parameter for generating the noise baseline level.

[0058] S2.4, based on the clustering results and the time-frequency domain energy distribution and change rate analysis of the noise source, the characteristics of each noise source are compared with the noise baseline level. Correspondingly, the noise floor level is calculated by the following generation logic: ;in, is the number of cluster categories, is the adaptive dynamic mapping function, and They represent the weighted functions of the clustering matrix and the noise change rate respectively (specifically in the form of polynomial operations), and the contribution ratio is automatically adjusted according to the noise type.

[0059] The noise floor level will determine the device's sensitivity adjustment and dynamic threshold setting for voice control in different environments, further optimizing the device's intelligent response capabilities.

[0060] Step S2 effectively generates a noise reference level adapted to different typical scenarios by integrating feature clustering results, time-frequency domain energy analysis, and an adaptive dynamic mapping model of noise change rate. First, based on the clustering processing of the sound source characteristic vector, a multidimensional noise feature space is established, and the complex time-frequency characteristics of each noise category are obtained through deep mapping. Then, through the energy change calculation in the time domain and frequency domain, combined with the volatility and change rate analysis of the noise, the real-time changes of the noise environment are dynamically captured to ensure that the characteristics of the noise source can accurately reflect its changing trend. Finally, the noise reference level is generated through an adaptive mapping function, thereby providing an accurate basis for adjusting the voice control sensitivity of the device. This processing process not only breaks through the traditional simple analysis method, but also improves the accuracy of the noise level and the environmental adaptability of the device through complex multidimensional data fusion.

[0061] In the application of intelligent voice devices, how to adaptively adjust the time window and pickup range of voice recognition according to the characteristics of environmental noise is the key to improving the efficiency and accuracy of voice recognition. In a home or office environment, background noise and environmental changes have a significant impact on the performance of the voice recognition system. Especially in a variable noise environment, a fixed window duration and pickup range can easily lead to frequent misrecognition or missing of valid commands by the system. Therefore, dynamic adjustment of sensitivity not only helps to improve recognition accuracy, but also enhances the device's ability to adapt to different environments.

[0062] Step S3 includes the following contents:

[0063] In a home environment, especially when there are multiple sound sources at the same time, the device needs to have the ability to adapt to different sound sources and respond accordingly. Fluctuations in the noise environment (such as external traffic noise or conversations between family members) often affect the device's voice recognition accuracy. In this scenario, the static time window design cannot adapt to environmental changes. Therefore, the time window adjustment mechanism based on the noise baseline level and the noise change rate is particularly important. By dynamically adjusting the window length, the device can increase the response time in a lower noise environment and respond quickly in a noisy environment to avoid misrecognition.

[0064] S3.1, the goal of adaptive window optimization is to dynamically adjust the window length of speech recognition according to the current environmental noise level and fluctuation. Through this optimization process, the device can shorten the speech detection window in noisy environments to avoid recognizing unnecessary background noise; and in quiet environments, extend the window length to improve the accuracy and stability of speech recognition.

[0065] The duration of the speech detection window is dynamically adjusted according to the ambient noise level and change rate to adapt to different noise conditions.

[0066] Adjust window length The adjustment is based on the noise floor level and the noise change rate The interaction between them reflects the complexity and changing trend of environmental noise. The specific calculation process is as follows: ;in, : The current noise baseline level, reflecting the intensity of ambient noise. : The noise change rate at the current moment, reflecting the speed and frequency of noise changes. : It is used to characterize the impact of the noise change rate on the window duration. A larger noise change rate will result in a smaller window duration, ensuring a timely response to sudden noise. : As the noise floor level increases, the window duration gradually increases to meet the sensitivity adjustment needs in higher noise environments. : A small constant that prevents the window duration from being zero and ensures that the minimum window duration has practical significance.

[0067] At higher noise floor levels, the window length will increase accordingly to ensure the accuracy and response time of speech recognition.

[0068] In cases where the noise changes at a high rate, the window duration will be shortened to prevent false recognition and ensure a fast response.

[0069] This formula takes both the intensity and volatility of the noise into account, providing a more accurate and dynamic window duration adjustment mechanism.

[0070] S3.2, similar to the adjustment of the time window, the adjustment of the pickup range also needs to change dynamically according to the ambient noise intensity and noise fluctuations. An excessively large pickup range may cause the device to frequently capture irrelevant background noise, reducing the accuracy and efficiency of speech recognition. On the other hand, a pickup range that is too small may lead to missed commands, especially at long distances or at low volumes. Therefore, based on the adjustment mechanism of the noise baseline level and the noise change rate, the pickup range can be effectively adjusted dynamically according to changes in the noise environment, ensuring that the device can reduce the impact of background noise in different environments and ensure the capture of voice signals when necessary.

[0071] Pickup range The adjustment is based on the interaction between the noise floor level and the noise change rate, aiming to optimize the sensitivity and coverage of the microphone array to ensure the accuracy of sound source identification and response efficiency. The specific calculation process is as follows: ;in, :The impact of noise baseline level on the pickup range. As the noise intensity increases, the pickup range gradually decreases to avoid misidentification and environmental noise interference. : The effect of noise change rate on the pickup range. When the noise change rate is high, the pickup range will be significantly reduced, ensuring that the device can better recognize voice commands in a high dynamic noise environment. It is a constant used to smooth the influence of noise change rate. Setting its value can adjust the influence of change rate on the pickup range.

[0072] Step S3 dynamically adjusts the voice detection window duration and pickup range by combining the noise reference level and the noise change rate, and using the adaptive window optimization and pickup range adjustment algorithm. In this process, the noise reference level reflects the overall intensity of the current environmental noise, and the noise change rate measures the volatility and change trend of the noise. According to these two parameters, the device can shorten the voice detection window duration and limit the pickup range in an environment with strong or rapidly changing noise, thereby effectively avoiding noise interference and false triggering. In an environment with low noise or stable changes, the device can extend the window duration and expand the pickup range to ensure that voice commands at a longer distance or lower volume can also be accurately captured and responded to. Through this sensitivity adjustment based on the characteristics of environmental noise, the device can dynamically adjust the sensitivity of voice recognition in different usage scenarios to ensure recognition accuracy and optimize the user experience.

[0073] Embodiment 2: Figure 2 The present invention provides a control system of an acoustic early childhood education machine, comprising: a sound source detection module, a noise assessment module and a sensitivity control module;

[0074] Sound source detection module: Based on a multi-microphone array, combined with directional filtering and spectrum segmentation technology, it obtains multi-dimensional audio features in the environment and generates a sound source characteristic vector containing sound source distance, direction and frequency information; the sound source characteristic vector is transmitted to the subsequent noise assessment module;

[0075] Noise assessment module: Through the sound source characteristic vector, feature clustering and time-frequency domain energy analysis, combined with adaptive dynamic mapping of noise change rate, the noise volatility and change trend of the current environment are evaluated, and the noise reference level suitable for different typical scenarios is generated; the noise reference level is transmitted to the sensitivity control module;

[0076] Sensitivity control module: dynamically adjusts the voice detection window duration and pickup range based on the noise baseline level and change rate.

[0077] The above formulas are all dimensionless and numerical calculations. 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.

[0078] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0079] It should be noted that, in this article, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including a..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0080] 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. A control method for an acoustic early childhood education machine, characterized in that: Includes steps: S1, based on a multi-microphone array and combined with directional filtering and spectrum segmentation technology, obtains the multi-dimensional audio characteristics of the sound source environment and generates a sound source characteristic vector containing the sound source distance, direction, and frequency information; S2, based on the sound source characteristic vector, through feature clustering and time-frequency domain energy analysis, combined with adaptive dynamic mapping of noise change rate, evaluates the volatility and change trend of the current environmental noise, and generates noise reference levels suitable for different typical scenarios; S3, dynamically adjusts the speech detection window duration and pickup range by combining the noise baseline level and change rate; Step S2 includes the following contents: S2.1, first, cluster the sound source characteristic vectors to obtain the cluster category to which each sound source belongs; on this basis, the adaptive mapping model of time-frequency domain distribution is used to perform deep mapping on the clustering results, establish the multidimensional feature space of each noise source, and set each noise source category of the clustering results as , and its mapping matrix in the feature space is ; The generation formula of the mapping matrix is: ;in, For the The audio feature vector of cluster samples, are the corresponding time and frequency dimension features, represents the tensor product of a vector and a matrix, For the The number of class samples; S2.2, after feature clustering, extract the time domain and frequency domain features of each noise source category; evaluate the volatility and distribution trend of noise through the instantaneous energy change in the time domain and the spectrum drift characteristics in the frequency domain; set the time domain energy and frequency domain energy The calculation formula is as follows: ; ;in, is the time signal at time The amplitude of The spectrum signal at frequency The amplitude of and Represents the number of sampling points in time and frequency respectively, Represents the sampling point index of the time domain signal, which is used for time domain energy calculation. Represents the frequency point index of the frequency domain signal, which is used for frequency domain energy calculation; by combining time domain energy and frequency domain energy, a multidimensional energy function is constructed , and dynamically track the noise source: ;in, and is the dynamic adaptive scaling factor; S2.3, after extracting the time-frequency domain features, calculate the volatility and change rate of each noise source category; through the incremental analysis of the time-frequency features, define the noise change rate as the incremental ratio of the noise energy, and the specific formula is: ;in, Indicates the rate of change of noise energy; S2.4, based on the clustering results and the time-frequency domain energy distribution and change rate analysis of the noise source, the characteristics of each noise source are compared with the noise baseline level. Correspondingly, the noise floor level is calculated by the following generation logic: ;in, is the number of cluster categories, is the adaptive dynamic mapping function, and Represent the weighting functions of the clustering matrix and the noise change rate respectively.

2. The control method of an acoustic early education machine according to claim 1, characterized in that: Step S1 includes the following contents: S1.1, using multiple microphones built into the device to form an array, assuming that the spatial position of each microphone is known, each microphone records the arrival time of sound waves in different directions to form an initial sound collection matrix based on spatial distribution, where the rows of the matrix represent microphone channels and the columns represent the audio signal values ​​that change over time; S1.2, based on the time difference of microphone signal arrival, a spatial sound source localization model is constructed, and the cross-correlation between signals is calculated using the generalized cross-power spectrum method to obtain the time difference matrix of each microphone pair; S1.3, based on the signal after directional filtering, the average distance between the sound source and the microphone array is estimated using the attenuation model; S1.4, perform Fourier transform on the signal after directional filtering to obtain spectrum information; use spectrum segmentation method to decompose the signal into multiple frequency bands, and screen out the main frequency peaks to obtain frequency energy concentration; extract the main frequency peaks and frequency band energy proportions of the sound source through the multiple frequency bands segmented to form characteristic frequencies; S1.5, comprehensively analyzing the processing results, combining the average distance, directional angle and pitch angle, characteristic frequency, and frequency concentration parameters of the sound source into a sound source characteristic vector.

3. The control method of an acoustic early education machine according to claim 1, characterized in that: Step S3 includes the following contents: S3.1, adjust the window length The adjustment is based on the interaction between the noise baseline level and the noise change rate, reflecting the complexity and change trend of environmental noise. The specific calculation process is as follows: ;in, Indicates the noise baseline level at the current moment; Indicates the noise change rate at the current moment; is a small constant to prevent the window length from being zero.

4. The control method of an acoustic early education machine according to claim 3, characterized in that: Step S3 also includes the following contents: S3.2, pickup range The adjustment is based on the interaction between the noise floor level and the noise change rate, aiming to optimize the sensitivity and coverage of the microphone array to ensure the accuracy of sound source identification and response efficiency. The specific calculation process is as follows: ;in, is a constant used to smooth the effects of the rate of change of noise.

5. A control system of an acoustic early childhood education machine, used to implement a control method of an acoustic early childhood education machine according to any one of claims 1 to 4, characterized in that: include: Sound source detection module, noise assessment module and sensitivity control module; Sound source detection module: Based on a multi-microphone array, combined with directional filtering and spectrum segmentation technology, it obtains multi-dimensional audio features in the environment and generates a sound source characteristic vector containing sound source distance, direction and frequency information; the sound source characteristic vector is transmitted to the subsequent noise assessment module; Noise assessment module: Through the sound source characteristic vector, feature clustering and time-frequency domain energy analysis, combined with adaptive dynamic mapping of noise change rate, it evaluates the noise volatility and change trend of the current environment and generates noise benchmark levels suitable for different typical scenarios; passing the noise reference level to the sensitivity control module; Sensitivity control module: dynamically adjusts the voice detection window duration and pickup range based on the noise baseline level and change rate.

Images