Intelligent control system and method of loudspeaker
The intelligent control system uses deep learning and neural networks to analyze audio signals, dynamically adjusting power output in loudspeakers, thereby improving audio quality stability and clarity by matching power requirements.
Patent Information
- Application Number
- CN202411203126.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-29
- Publication Date
- 2024-12-03
AI Technical Summary
Existing loudspeaker systems struggle to dynamically adjust power output to match the varying power requirements of different audio signals, leading to instability and poor audio quality.
An intelligent control system that utilizes frequency domain and waveform analysis through deep learning and neural networks to adjust amplifier power supply voltage in real time, incorporating Fourier transforms, convolutional neural networks, and linear interpolation modulation to optimize power output.
Enhances audio quality stability and clarity by dynamically adapting power output to meet the demands of varying audio conditions, reducing distortion and improving sound realism.
Smart Images

Figure CN119065303A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of intelligent control, and more specifically, to an intelligent control system and method for a speaker. Background Art
[0002] Different types of audio signals have different power requirements. For example, in the soft part of the music, the required power is low, and unnecessary power loss needs to be reduced; when the music enters the strong part such as the heavy bass, the required power is high, and the power needs to be increased to better restore the details and dynamic range of the audio signal. Without dynamic voltage adjustment, some parts of the audio signal will have insufficient or excessive power, which will affect the stability of the audio quality. In order to ensure that the speaker can output with appropriate power under various audio signal conditions, thereby optimizing the stability of the audio quality, making the sound clearer and more realistic, and avoiding distortion, broken sound and other problems, an intelligent control solution for the speaker is expected. Summary of the invention
[0003] In view of the shortcomings of the prior art, the present application provides an intelligent control system and method for a loudspeaker.
[0004] According to one aspect of the present application, there is provided an intelligent control system for a loudspeaker, comprising:
[0005] An audio signal acquisition module, used to obtain the audio signal played by the speaker;
[0006] An audio signal frequency domain feature encoding module, used for performing frequency domain feature encoding on the audio signal to obtain an audio signal frequency domain feature vector;
[0007] An audio signal waveform encoding module, used for performing feature encoding on the waveform of the audio signal to obtain an audio signal waveform emphasis feature vector;
[0008] An audio signal frequency domain feature optimization module, used for performing feature abnormality stacking processing based on linear interpolation modulation on the audio signal frequency domain feature vector based on the audio signal waveform emphasis feature vector to obtain an optimized audio signal frequency domain feature vector;
[0009] An audio signal feature fusion module, used for fusing the optimized audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector to obtain an audio signal feature vector;
[0010] The power supply voltage adjustment direction result generating module is used to obtain the adjustment direction of the amplifier output power supply voltage based on the audio signal feature vector.
[0011] According to another aspect of the present application, there is provided a method for intelligently controlling a speaker, comprising:
[0012] Get the audio signal played by the speaker;
[0013] Performing frequency domain feature encoding on the audio signal to obtain a frequency domain feature vector of the audio signal;
[0014] Performing feature encoding on the waveform graph of the audio signal to obtain an audio signal waveform emphasis feature vector;
[0015] Based on the audio signal waveform emphasis feature vector, the audio signal frequency domain feature vector is subjected to feature abnormal stacking processing based on linear interpolation modulation to obtain an optimized audio signal frequency domain feature vector;
[0016] fusing the optimized audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector to obtain an audio signal feature vector;
[0017] Based on the audio signal characteristic vector, an adjustment direction of the amplifier output power supply voltage is obtained.
[0018] This application has significant technical effects due to the adoption of the above technical solutions:
[0019] The intelligent control system and method of the loudspeaker provided by the present application can better understand the dynamic changes of the audio signal by analyzing the frequency domain and waveform characteristics of the audio signal, so as to adjust the output power supply voltage of the loudspeaker amplifier in real time. In this way, by adjusting the amplifier power supply voltage, dynamic adjustment of the power output is achieved, so as to meet the power requirements under different audio signal conditions, which is conducive to improving the stability and clarity of the sound quality. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.
[0021] Figure 1 The figure is a flow chart of a method for intelligently controlling a speaker according to an embodiment of the present application.
[0022] Figure 2 Schematic diagram of data flow of the intelligent control method of the speaker according to an embodiment of the present application.
[0023] Figure 3 This is a flowchart of performing frequency domain feature encoding on the audio signal to obtain a frequency domain feature vector of the audio signal in the intelligent control method of the speaker according to an embodiment of the present application.
[0024] Figure 4 The present invention is a flowchart of performing feature encoding on the waveform diagram of the audio signal to obtain an audio signal waveform emphasis feature vector in the intelligent control method of the speaker according to an embodiment of the present application.
[0025] Figure 5 4 is a system block diagram of an intelligent control system of a speaker according to an embodiment of the present application. DETAILED DESCRIPTION
[0026] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.
[0027] Different types of audio signals have great differences in power requirements. The power required for the soft part of the music is relatively low, and unnecessary power loss should be reduced at this time; when the music enters the strong part such as the heavy bass, the required power is higher, and the power needs to be increased to better restore the details and dynamic range of the audio signal. If the voltage is not adjusted dynamically, some audio signals will have insufficient or excessive power, which will affect the stability of the audio quality. In order to enable the speaker to output with appropriate power under various audio signal conditions, to optimize the stability of the audio quality, to make the sound clearer and more realistic, and to prevent distortion, broken sound and other problems, it is hoped that there will be an intelligent control solution for the speaker.
[0028] In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also shown a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields. The development of deep learning and neural networks provides new solutions and solutions for the intelligent control of speakers.
[0029] Figure 1 The figure is a flow chart of a method for intelligently controlling a speaker according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the intelligent control method of the speaker according to the embodiment of the present application. Figure 1 and Figure 2As shown, the intelligent control method of the speaker according to the embodiment of the present application includes: S110, obtaining the audio signal played by the speaker; S120, performing frequency domain feature encoding on the audio signal to obtain an audio signal frequency domain feature vector; S130, performing feature encoding on the waveform graph of the audio signal to obtain an audio signal waveform emphasis feature vector; S140, performing feature anomaly stacking processing based on linear interpolation modulation on the audio signal frequency domain feature vector based on the audio signal waveform emphasis feature vector to obtain an optimized audio signal frequency domain feature vector; S150, fusing the optimized audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector to obtain an audio signal feature vector; S160, based on the audio signal feature vector, obtaining the adjustment direction of the amplifier output power supply voltage.
[0030] In step S110, the audio signal played by the speaker is obtained. It should be understood that the audio signal is usually a time series, indicating the change of the sound wave in time. Specifically, the audio signal is composed of multiple frequency components, and the components of different frequencies will affect the sound quality and the characteristics of the sound (such as bass, midrange, and treble). The energy distribution of different frequency components directly affects the power output. For example, low-frequency signals usually require higher power to drive the speaker to produce sufficient bass effect. By analyzing the frequency domain characteristics of the audio signal, the amplifier can increase the power output in the heavy bass part, while maintaining a lower power in the treble or midrange part. The change (waveform) of the audio signal in time can reflect the dynamic characteristics of the music, including instantaneous volume changes, the ups and downs of notes, etc., which will also affect the working state of the speaker and the required power output. In general, the audio signal played by the speaker contains a variety of information such as frequency and time domain changes, which provides a basis for adjusting the output power supply voltage of the amplifier. By analyzing this information in real time, the amplifier can dynamically adapt to the changes in the audio signal, thereby optimizing the power output, thereby improving the stability and clarity of the sound quality.
[0031] In step S120, the audio signal is encoded in frequency domain to obtain a frequency domain feature vector of the audio signal. Specifically, Figure 3 The flowchart is a method for intelligently controlling a loudspeaker according to an embodiment of the present application, in which the frequency domain feature encoding is performed on the audio signal to obtain a frequency domain feature vector of the audio signal. Figure 3 As shown, frequency domain feature encoding is performed on the audio signal to obtain a frequency domain feature vector of the audio signal, including: S121, Fourier transform is performed on the audio signal to obtain multiple frequency domain statistical features of the audio signal; S122, the multiple frequency domain statistical features of the audio signal are arranged into a one-dimensional frequency domain input vector and then passed through a sequence encoder to obtain the frequency domain feature vector of the audio signal.
[0032] The audio signal is composed of multiple frequency components superimposed. In order to extract the frequency components of the audio signal and its energy distribution at different frequencies, the technical solution of the present application needs to perform Fourier transform on the audio signal to obtain multiple audio signal frequency domain statistical features. Fourier transform is a mathematical transformation method that aims to convert time domain signals into frequency domain signals. Its basic idea is that any periodic or non-periodic signal can be represented as a superposition of a set of sine waves. By performing Fourier transform on the audio signal, the frequency domain characteristics of the audio signal can be identified. Through the frequency domain characteristics, important frequency components in the audio signal, such as bass, midrange and treble, can be identified, thereby providing a basis for subsequent power adjustment.
[0033] In order to integrate multiple audio signal frequency domain statistical features into a compact feature representation and capture the relationship and dependency between different audio signal frequency domain statistical features, the technical solution of the present application arranges the multiple audio signal frequency domain statistical features into a one-dimensional frequency domain input vector and then passes through a sequence encoder to obtain the audio signal frequency domain feature vector. The sequence encoder can automatically learn and extract the complex relationship between frequency domain features and identify the most important part for power supply voltage adjustment. In short, through the extracted audio signal frequency domain feature vector, the model can better understand the dynamic changes of the signal, thereby achieving more flexible power supply voltage adjustment. In the present application, the sequence encoder is a convolutional neural network model including a fully connected layer and a one-dimensional convolutional layer. The one-dimensional convolutional layer can effectively extract local features in the input vector and identify the local correlation between frequency domain features. For example, some frequency components may show similar change patterns in time, and the convolutional layer can capture these patterns. Through multiple layers of convolution and fully connected layers, the sequence encoder can construct a hierarchical feature representation, from low-level simple features to high-level complex features, and enhance the expression ability of information in the one-dimensional frequency domain input vector.
[0034] In step S130, feature encoding is performed on the waveform of the audio signal to obtain an audio signal waveform emphasis feature vector. Specifically, Figure 4 The present invention is a flowchart of performing feature encoding on the waveform of the audio signal to obtain an audio signal waveform emphasis feature vector in the intelligent control method of the speaker according to the embodiment of the present application. Figure 4 As shown, the waveform graph of the audio signal is feature encoded to obtain an audio signal waveform emphasis feature vector, including: S131, passing the waveform graph of the audio signal through a waveform feature encoder to obtain an audio signal waveform feature matrix; S132, passing the audio signal waveform feature matrix through a waveform feature emphasizer to obtain an audio signal waveform emphasis feature matrix; S133, expanding the audio signal waveform emphasis feature matrix into the audio signal waveform emphasis feature vector.
[0035] The waveform graph of an audio signal is a visual representation of the changes in the audio signal in the time domain. Specifically, the vertical axis of the waveform graph represents the amplitude of the audio signal (i.e., sound pressure level), showing the changes in the intensity of the sound; the horizontal axis of the waveform graph represents time, showing the changes in the audio signal over time, and providing the timing characteristics of the signal. In order to extract the time domain characteristics of the audio signal, such as the instantaneous amplitude, the shape characteristics of the waveform (such as peaks, troughs, etc.), it is necessary to pass the waveform graph of the audio signal through a waveform feature encoder to obtain an audio signal waveform feature matrix. After being processed by the waveform feature encoder, the instantaneous amplitude and change pattern of the audio signal can be identified, so that the amplifier can more accurately adjust the power supply voltage to optimize the output volume and sound quality. The waveform feature encoder in this application is a convolutional neural network model using a hole convolution kernel. Dilated Convolution introduces holes in the convolution operation (i.e. inserting gaps between convolution kernel elements), so that the convolution kernel can cover a larger receptive field without increasing the amount of computation. Its main advantages include: a. Dilated convolution can expand the receptive field without increasing the number of parameters, so that the model can capture features over a longer time range; b. By adjusting the dilation rate, features can be extracted at different scales, which helps to capture changes in different frequency components in audio signals.
[0036] In the waveform feature matrix of the audio signal, some features are more important to the final power supply voltage adjustment. In order to emphasize these key features and improve the performance of the speaker control model, it is necessary to pass the audio signal waveform feature matrix through a waveform feature emphasizer to obtain an audio signal waveform emphasized feature matrix. The waveform feature emphasizer in this application is a convolutional neural network model using a spatial attention mechanism. The spatial attention mechanism highlights the important features in the original waveform feature matrix and suppresses unimportant features by weighting different areas of the input feature matrix. This mechanism enables the model to focus on more relevant information. By emphasizing important features, we can better focus on the key parts of the original audio signal, thereby improving processing accuracy.
[0037] In order to make the audio signal waveform feature processing more efficient, it is necessary to expand the audio signal waveform emphasized feature matrix into an audio signal waveform emphasized feature vector. It should be understood that expanding the audio signal waveform feature matrix into a feature vector can effectively reduce the feature dimension and make the feature data more concise, which helps to reduce the complexity of subsequent processing. And the expanded feature vector can be more easily integrated with other features, thereby enhancing the ability to understand complex audio signals.
[0038] In step S140, based on the audio signal waveform emphasized feature vector, the audio signal frequency domain feature vector is subjected to feature abnormal stacking processing based on linear interpolation modulation to obtain an optimized audio signal frequency domain feature vector. It should be understood that the audio signal frequency domain feature vector is extracted from the original audio time domain signal through Fourier transform, while the audio signal waveform emphasized feature vector is directly extracted from the original audio time domain signal. In the encoding process of obtaining the two feature vectors, the emphasis on processing the audio signal is different, which will lead to differences in the details of the extracted audio features. In particular, in signal processing, there is a trade-off between time resolution and frequency resolution, namely the Heisenberg uncertainty principle. In frequency domain feature analysis, if a finer frequency resolution analysis is performed on the audio signal, the time resolution may be sacrificed, and vice versa. This trade-off may lead to inconsistencies in details between the audio signal frequency domain feature and the audio waveform feature. That is, there is a fine-grained inconsistency in feature distribution between the audio signal frequency domain feature vector and the audio signal waveform emphasized feature vector, which will cause the fused feature vector to have abnormal stacking on the boundary of the feature manifold. Therefore, in the technical solution of the present application, it is necessary to perform feature anomaly stacking processing based on linear interpolation modulation on the audio signal frequency domain feature vector based on the audio signal waveform emphasis feature vector to obtain an optimized audio signal frequency domain feature vector.
[0039] Specifically, in an embodiment of the present application, based on the audio signal waveform emphasized feature vector, the audio signal frequency domain feature vector is subjected to feature anomaly stacking processing based on linear interpolation modulation to obtain an optimized audio signal frequency domain feature vector, including: calculating the positional difference between the audio signal frequency domain feature vector and the audio signal waveform emphasized feature vector to obtain an audio signal shift positional difference representation vector; after cascading the audio signal frequency domain feature vector and the audio signal waveform emphasized feature vector into an audio signal joint feature vector, the audio signal joint feature vector is input into a one-dimensional convolution layer to obtain an audio signal shift convolution coding representation vector; after linear embedding modulation encoding the audio signal frequency domain feature vector using a first embedding matrix into an audio signal frequency domain feature linear modulation vector, the audio signal is calculated. The method comprises the following steps: performing a positional difference between a frequency domain feature linear modulation vector of an audio signal and a waveform emphasis feature vector of the audio signal to obtain a shifted modulation differential representation vector of the audio signal; taking the audio signal shifted positional difference representation vector as a value vector, the audio signal shifted convolutional coding representation vector and the audio signal shifted modulation differential representation vector as a query vector and a key vector, inputting the audio signal shifted positional difference representation vector, the audio signal shifted convolutional coding representation vector and the audio signal shifted modulation differential representation vector into a shift corrector based on an emulated converter structure to obtain an audio signal multi-scale hierarchical fusion shifted modulation weight vector; calculating the positional point multiplication between the audio signal multi-scale hierarchical fusion shifted modulation weight vector and the audio signal frequency domain feature vector to obtain the optimized audio signal frequency domain feature vector.
[0040] In an embodiment of the present application, another expression of performing feature abnormal stacking processing based on linear interpolation modulation on the audio signal frequency domain feature vector based on the audio signal waveform emphasis feature vector to obtain an optimized audio signal frequency domain feature vector may be: the audio signal waveform emphasis feature vector and the audio signal frequency domain feature vector are processed according to the following formula to obtain an optimized audio signal frequency domain feature vector; wherein the formula is:
[0041]
[0042] V 2d =Conv 1D ([V 1 ; V 2 ])
[0043]
[0044] V 1 '=V 1 ⊙V d
[0045] Among them, V1 represents the frequency domain feature vector of the audio signal, V 2 represents the audio signal waveform emphasized feature vector, Indicates point-by-point subtraction, V 1d Indicates that the audio signal is shifted by position difference to represent the vector, [;] indicates the cascade operation, Conv 1 D(·) represents one-dimensional convolutional coding, V 2d represents the audio signal shift convolutional coding representation vector, W 1 represents the first embedding matrix, b 1 represents the first modulation bias vector, V 3d A vector representing the shift modulated difference representation of the audio signal, represents vector multiplication, Softmax represents the normalized exponential function, T represents the transposed vector of the feature vector, d represents the length of the feature vector, Relu represents the activation function, V d represents the multi-scale hierarchical fusion shift modulation weight vector of the audio signal, ⊙ represents the position point multiplication, V 1 ' represents the optimized audio signal frequency domain feature vector.
[0046] More specifically, in an embodiment of the present application, after the audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector are cascaded into an audio signal joint feature vector, the audio signal joint feature vector is input into a one-dimensional convolution layer to obtain an audio signal shift convolution coding representation vector, including: creating a C++ file; writing a function in the C++ file to implement feature vector cascading and feature vector one-dimensional convolution coding; and using a C++ compiler to compile the C++ file to generate an executable file.
[0047] In an embodiment of the present application, after the audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector are cascaded into an audio signal joint feature vector, the audio signal joint feature vector is input into a one-dimensional convolution layer to obtain a code for obtaining an audio signal shift convolution coding representation vector. The implementation method is:
[0048]
[0049]
[0050]
[0051] The advantages of using C++ to deploy and implement "after concatenating the audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector into an audio signal joint feature vector, the audio signal joint feature vector is input into a one-dimensional convolution layer to obtain an audio signal shift convolution coding representation vector" include: a. Code written in C++ can be compiled and run on a variety of operating systems, including Windows, Linux, and macOS, which provides flexibility for deployment; b. C++ has rich audio processing and machine learning libraries, such as FFTW (Fast Fourier Transform Library), OpenCV (Open Source Computer Vision Library), and Dlib (Machine Learning Library), which can help developers quickly implement complex audio processing functions; c. C++ allows developers to manually manage memory, which helps avoid memory leaks and optimize memory usage, especially when processing large amounts of data; d. Compared with solutions that rely on specific platforms or languages, using C++ can reduce licensing fees and runtime costs; e. C++ code is easy to extend and maintain, which is very useful for situations where new features may need to be added or existing features improved as the project evolves.
[0052] It should be understood that in the technical solution of the present application, in the process of fusing the audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector, a feature anomaly stacking processing method based on linear interpolation modulation is proposed to address the problem of fine-grained inconsistency in feature distribution that arises in the process of fusing the audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector, aiming to reduce the abnormal stacking phenomenon of the fused feature vector on the boundary of the feature manifold, so as to improve the robustness and classification performance of the model.
[0053] Specifically, the method first calculates the element-by-element difference between the audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector to obtain the audio signal shift position differential representation vector, thereby capturing the difference between the two feature vectors. Then, the audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector are cascaded into an audio signal joint feature vector and input into a one-dimensional convolution layer to generate an audio signal shift convolution coding representation vector. That is, the deep-level shift implicit information between the audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector is captured by one-dimensional convolution coding. At the same time, the audio signal frequency domain feature vector is linearly embedded and modulated to be an audio signal frequency domain feature linear modulation vector using the first embedding matrix, and the element-by-element difference between the audio signal frequency domain feature linear modulation vector and the audio signal waveform emphasis feature vector is calculated to obtain the audio signal shift modulation differential representation vector. The audio signal shift modulation differential representation vector is used to represent the position expression of the shift information of the audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector in the linear space. That is, in the technical solution of the present application, the audio signal shift position differential representation vector, the audio signal shift convolutional coding representation vector and the audio signal shift modulation differential representation vector respectively represent the explicit expression of the fine-grained distribution difference shift information between the audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector in different feature spaces.
[0054] Next, the audio signal shift position differential representation vector is used as the value vector, the audio signal shift convolutional coding representation vector and the audio signal shift modulation differential representation vector are used as the query vector and the key vector respectively, and these vectors are input into the shift corrector based on the simulated transformer structure to generate the audio signal multi-scale hierarchical fusion shift modulation weight vector. In this way, the attention mechanism is used to evaluate the importance of features of different scales, that is, the attention mechanism is used to simulate the hierarchical precipitation expression of the fine-grained difference features between the audio signal frequency domain feature vector and the audio signal waveform emphasizing feature vector in different feature spaces. Finally, the element-by-element product of the audio signal multi-scale hierarchical fusion shift modulation weight vector and the audio signal frequency domain feature vector is calculated to obtain the optimized audio signal frequency domain feature vector, that is, the weighted operation is used to assign weights to each element in the audio signal frequency domain feature vector, which reduces the influence of local disturbances and improves the quality of the feature vector.
[0055] In summary, in the method, the audio signal shift position differential representation vector, the audio signal shift convolutional coding representation vector and the audio signal shift modulation differential representation vector respectively represent the explicit expression of the fine-grained distribution difference shift information between the audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector in different feature spaces. The generation of the multi-scale hierarchical fusion shift modulation weight vector of the audio signal uses the attention mechanism to evaluate the importance of features of different scales and provide weight information for feature shift correction. Ultimately, the generation of the optimized audio signal frequency domain feature vector reduces the impact of disturbances, improves the robustness of the model to local disturbances in the input data, enhances the model's adaptability to noisy or variant data sets, improves the fusion quality of the feature vector, and improves the overall classification performance and training efficiency of the model.
[0056] In step S150, the optimized audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector are fused to obtain the audio signal feature vector. It should be understood that the optimized audio signal frequency domain feature vector reflects the frequency spectrum information of the audio signal, including frequency components, amplitude, etc.; the audio signal waveform emphasis feature vector focuses on the changes in the time domain signal, reflects the dynamic characteristics and instantaneous changes of the signal, and can capture information such as instantaneous volume changes and transient responses; by fusing the optimized audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector, the spectrum and time domain information of the signal can be comprehensively considered, thereby improving the expression ability of the audio signal characteristics and enhancing the understanding of complex audio signals. That is, the fused audio signal feature vector can more accurately reflect the characteristics of the audio signal, provide more comprehensive information, and enable the power supply voltage adjustment to more accurately respond to changes in the audio signal.
[0057] In step S160, based on the audio signal feature vector, the adjustment direction of the amplifier output power supply voltage is obtained. Specifically, in an embodiment of the present application, based on the audio signal feature vector, the adjustment direction of the amplifier output power supply voltage is obtained, including: passing the audio signal feature vector through an amplifier output power supply voltage adjuster based on a classifier to obtain a classification result, and the classification result is used to indicate that the amplifier output power supply voltage value should be increased, decreased or unchanged. It should be understood that the characteristics of audio signals are often complex. In order to map these complex features to a simple power supply voltage decision space (increase, decrease or remain unchanged), it is necessary to pass the audio signal feature vector through an amplifier output power supply voltage adjuster based on a classifier. A classifier is a machine learning model for classifying an input audio signal feature vector into different categories. It can establish a mapping relationship between input features and output categories by learning an existing data set. By adjusting the power supply voltage according to the classification result, the output power can be indirectly adjusted to optimize the sound quality, reduce distortion and noise, and thus improve the overall audio performance.
[0058] In an embodiment of the present application, the audio signal feature vector is passed through an amplifier output power supply voltage adjuster based on a classifier to obtain a classification result, and the classification result is used to indicate that the amplifier output power supply voltage value should be increased, decreased, or remain unchanged. One possible implementation method may be: using the fully connected layer of the classifier to fully connect the audio signal feature vector to obtain an audio signal fully connected encoded feature vector; inputting the audio signal fully connected encoded feature vector into the Softmax classification function of the classifier to obtain probability values belonging to each classification label in the audio signal feature vector, and the classification labels include those for indicating that the amplifier output power supply voltage value should increase, for indicating that the amplifier output power supply voltage value should decrease, or for indicating that the amplifier output power supply voltage value should remain unchanged; and determining the classification label corresponding to the largest of the probability values as the classification result.
[0059] In summary, the intelligent control method of the speaker based on the embodiment of the present application is explained, which can better understand the dynamic changes of the audio signal by analyzing the frequency domain and waveform characteristics of the audio signal, so as to adjust the output power supply voltage of the speaker amplifier in real time. In this way, by adjusting the amplifier power supply voltage, dynamic adjustment of the power output is achieved, thereby meeting the power requirements under different audio signal conditions, which is conducive to improving the stability and clarity of the sound quality.
[0060] Figure 5 FIG. 1 is a system block diagram of an intelligent control system for a loudspeaker according to an embodiment of the present application. Figure 5 As shown, the intelligent control system 100 of the speaker according to the embodiment of the present application includes: an audio signal acquisition module 110, which is used to obtain the audio signal played by the speaker; an audio signal frequency domain feature encoding module 120, which is used to perform frequency domain feature encoding on the audio signal to obtain an audio signal frequency domain feature vector; an audio signal waveform encoding module 130, which is used to perform feature encoding on the waveform of the audio signal to obtain an audio signal waveform emphasis feature vector; an audio signal frequency domain feature optimization module 140, which is used to perform feature anomaly stacking processing based on linear interpolation modulation on the audio signal frequency domain feature vector based on the audio signal waveform emphasis feature vector to obtain an optimized audio signal frequency domain feature vector; an audio signal feature fusion module 150, which is used to fuse the optimized audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector to obtain an audio signal feature vector; a power supply voltage adjustment direction result generation module 160, which is used to obtain the adjustment direction of the amplifier output power supply voltage based on the audio signal feature vector.
[0061] Here, those skilled in the art will appreciate that the specific functions and operations of the various units and modules in the above-mentioned speaker intelligent control system 100 have been described in detail above. Figures 1 to 4 The description of the intelligent control method of the speaker has been introduced in detail, and therefore, its repeated description will be omitted.
[0062] In summary, the intelligent control system 100 of the loudspeaker according to the embodiment of the present application is explained, which can better understand the dynamic changes of the audio signal by analyzing the frequency domain and waveform characteristics of the audio signal, so as to adjust the output power supply voltage of the loudspeaker amplifier in real time. In this way, by adjusting the amplifier power supply voltage, dynamic adjustment of the power output is achieved, thereby meeting the power requirements under different audio signal conditions, which is conducive to improving the stability and clarity of the sound quality.
[0063] As described above, the intelligent control system 100 of the speaker according to the embodiment of the present application can be implemented in various wireless terminals, such as a server for intelligent control of the speaker. In one example, the intelligent control system 100 of the speaker according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the intelligent control system 100 of the speaker can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the intelligent control system 100 of the speaker can also be one of the many hardware modules of the wireless terminal.
[0064] Alternatively, in another example, the speaker's intelligent control system 100 and the wireless terminal may be separate devices, and the speaker's intelligent control system 100 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.
[0065] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details of the above application are only for the purpose of illustration and ease of understanding, rather than limitation, and the above details do not limit the present application to the use of the above specific details to be implemented.
Claims
1. An intelligent control method for a speaker, It is characterized in that include: Get the audio signal played by the speaker; Performing frequency domain feature encoding on the audio signal to obtain a frequency domain feature vector of the audio signal; Performing feature encoding on the waveform graph of the audio signal to obtain an audio signal waveform emphasis feature vector; Based on the audio signal waveform emphasis feature vector, the audio signal frequency domain feature vector is subjected to feature abnormal stacking processing based on linear interpolation modulation to obtain an optimized audio signal frequency domain feature vector; fusing the optimized audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector to obtain an audio signal feature vector; Based on the audio signal characteristic vector, an adjustment direction of the amplifier output power supply voltage is obtained.
2. The intelligent control method of the speaker according to claim 1, It is characterized in that Performing frequency domain feature encoding on the audio signal to obtain a frequency domain feature vector of the audio signal includes: Performing Fourier transform on the audio signal to obtain a plurality of audio signal frequency domain statistical features; The plurality of audio signal frequency domain statistical features are arranged into a one-dimensional frequency domain input vector and then passed through a sequence encoder to obtain the audio signal frequency domain feature vector.
3. The intelligent control method of the speaker according to claim 2, It is characterized in that Performing feature encoding on the waveform graph of the audio signal to obtain an audio signal waveform emphasis feature vector includes: Passing the waveform graph of the audio signal through a waveform graph feature encoder to obtain an audio signal waveform feature matrix; Passing the audio signal waveform feature matrix through a waveform feature emphasizer to obtain an audio signal waveform emphasized feature matrix; The audio signal waveform emphasis feature matrix is expanded into the audio signal waveform emphasis feature vector.
4. The intelligent control method of the speaker according to claim 3, It is characterized in that Based on the audio signal waveform emphasis feature vector, the audio signal frequency domain feature vector is subjected to feature abnormal stacking processing based on linear interpolation modulation to obtain an optimized audio signal frequency domain feature vector, including: Calculating the position difference between the audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector to obtain an audio signal shift position difference representation vector; After concatenating the audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector into an audio signal joint feature vector, the audio signal joint feature vector is input into a one-dimensional convolution layer to obtain an audio signal shift convolution coding representation vector; After performing linear embedding modulation encoding on the audio signal frequency domain feature vector using a first embedding matrix to obtain an audio signal frequency domain feature linear modulation vector, a position difference between the audio signal frequency domain feature linear modulation vector and the audio signal waveform emphasis feature vector is calculated to obtain an audio signal shift modulation difference representation vector; Taking the audio signal shifted position differential representation vector as a value vector, the audio signal shifted convolutional coding representation vector and the audio signal shifted modulation differential representation vector as a query vector and a key vector, the audio signal shifted position differential representation vector, the audio signal shifted convolutional coding representation vector and the audio signal shifted modulation differential representation vector are input into a shift corrector based on a simulated transformer structure to obtain an audio signal multi-scale hierarchical fusion shift modulation weight vector; The optimized audio signal frequency domain feature vector is obtained by calculating the point-by-point multiplication between the audio signal multi-scale hierarchical fusion shift modulation weight vector and the audio signal frequency domain feature vector.
5. The intelligent control method of the speaker according to claim 4, It is characterized in that After concatenating the audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector into an audio signal joint feature vector, the audio signal joint feature vector is input into a one-dimensional convolution layer to obtain an audio signal shift convolution coding representation vector, including: Create a C++ file; Writing functions in the C++ file to implement feature vector concatenation and feature vector one-dimensional convolution coding; The C++ file is compiled using a C++ compiler to generate an executable file.
6. The intelligent control method of the speaker according to claim 5, It is characterized in that The sequence encoder is a convolutional neural network model including a fully connected layer and a one-dimensional convolutional layer, the waveform feature encoder is a convolutional neural network model using a hole convolution kernel, and the waveform feature emphasizer is a convolutional neural network model using a spatial attention mechanism.
7. The intelligent control method of a loudspeaker according to claim 6, It is characterized in that Based on the audio signal feature vector, obtaining an adjustment direction of the amplifier output power supply voltage includes: The audio signal feature vector is passed through an amplifier output power supply voltage regulator based on a classifier to obtain a classification result, wherein the classification result is used to indicate whether the amplifier output power supply voltage value should be increased, decreased, or kept unchanged.
8. An intelligent control system for a loudspeaker, It is characterized in that include: An audio signal acquisition module, used to obtain the audio signal played by the speaker; An audio signal frequency domain feature encoding module, used for performing frequency domain feature encoding on the audio signal to obtain an audio signal frequency domain feature vector; An audio signal waveform encoding module, used for performing feature encoding on the waveform of the audio signal to obtain an audio signal waveform emphasis feature vector; An audio signal frequency domain feature optimization module, used for performing feature abnormality stacking processing based on linear interpolation modulation on the audio signal frequency domain feature vector based on the audio signal waveform emphasis feature vector to obtain an optimized audio signal frequency domain feature vector; An audio signal feature fusion module, used for fusing the optimized audio signal frequency domain feature vector and the audio signal waveform emphasis feature vector to obtain an audio signal feature vector; The power supply voltage adjustment direction result generating module is used to obtain the adjustment direction of the amplifier output power supply voltage based on the audio signal feature vector.
9. The intelligent control system of the loudspeaker according to claim 8, It is characterized in that The audio signal frequency domain feature encoding module comprises: A signal Fourier transform processing unit, used for performing Fourier transform on the audio signal to obtain a plurality of audio signal frequency domain statistical features; The audio signal frequency domain statistical feature sequence encoding unit is used to arrange the multiple audio signal frequency domain statistical features into a one-dimensional frequency domain input vector and then pass it through a sequence encoder to obtain the audio signal frequency domain feature vector.
10. The intelligent control system of the loudspeaker according to claim 9, It is characterized in that The power supply voltage adjustment direction result generating module is used to: pass the audio signal feature vector through the amplifier output power supply voltage adjuster based on the classifier to obtain a classification result, and the classification result is used to indicate whether the amplifier output power supply voltage value should be increased, decreased or kept unchanged.
Citation Information
Patent Citations
Noise reduction method and device for audio amplifier in standby state and electronic equipment
CN113593598A
Method for improving loudspeaker sound effect and related device
CN115708362A
Adaptive audio data acquisition and sound amplification data processing method
CN117119364A
Equipment life cycle management method and system
CN117273699A
Feature extraction method and apparatus based on time domain and frequency domain of speech signal, and echo cancellation method and apparatus
WO2023044962A1
