Signal Processing Method, Signal Processing Device, Terminal Device and Storage Medium

Abstract

This application is applicable to the field of signal processing technology, and provides a signal processing method, a signal processing device, a terminal device and a storage medium. The signal processing method includes: obtaining a signal to be processed; decomposing the signal to be processed into feature maps of N scales, where N is an integer greater than 1, and the scale of the signal to be processed is greater than the maximum scale among the N scales; inputting the feature maps of the N scales into a preset convolutional neural network to obtain a target feature map, and the scale of the target feature map is less than the minimum scale among the N scales. By this application, the computational amount and storage space required by the convolutional neural network can be reduced.

Description

Technical Field

[0001] This application belongs to the technical field of signal processing, and particularly relates to a signal processing method, a signal processing device, a terminal device, and a storage medium. Background Art

[0002] A convolutional neural network is a type of feedforward neural network that contains convolutional calculations and has a deep structure, and is one of the representative algorithms of deep learning.

[0003] A convolutional neural network is usually composed of several general forms of structural layers, such as an input layer, a convolutional layer, a pooling layer, and a fully connected layer. To deepen the convolutional neural network, related technologies usually stack the structural layers in the convolutional neural network, resulting in an increasing amount of computation and storage space required by the convolutional neural network. Summary of the Invention

[0004] Embodiments of this application provide a signal processing method, a signal processing device, a terminal device, and a storage medium to reduce the computation and storage space required by a convolutional neural network.

[0005] In a first aspect, embodiments of this application provide a signal processing method, and the signal processing method includes:

[0006] Obtain a signal to be processed;

[0007] Decompose the signal to be processed into feature maps of N scales, where N is an integer greater than 1, and the scale of the signal to be processed is greater than the maximum scale among the N scales;

[0008] Input the N-scale feature maps into a preset convolutional neural network to obtain a target feature map, and the scale of the target feature map is smaller than the minimum scale among the N scales.

[0009] In a second aspect, embodiments of this application provide a signal processing device, and the signal processing device includes:

[0010] A signal acquisition module, configured to obtain a signal to be processed;

[0011] A signal decomposition module, configured to decompose the signal to be processed into feature maps of N scales, where N is an integer greater than 1, and the scale of the signal to be processed is greater than the maximum scale among the N scales;

[0012] A feature map input module, configured to input the N-scale feature maps into a preset convolutional neural network to obtain a target feature map, and the scale of the target feature map is smaller than the minimum scale among the N scales.

[0013] In a third aspect, an embodiment of the present application provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the signal processing method described in the first aspect above are implemented.

[0014] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps of the signal processing method described in the first aspect above are implemented.

[0015] In a fifth aspect, an embodiment of the present application provides a computer program product, and when the computer program product runs on a terminal device, the terminal device is caused to execute the steps of the signal processing method described in the first aspect above.

[0016] As can be seen from the above, before inputting the signal to be processed into a preset convolutional neural network, the signal to be processed is first decomposed into feature maps of N scales. Since the scale of the signal to be processed is larger than the largest scale among the N scales, in this application, the large-scale signal to be processed is decomposed into multiple small-scale feature maps, and then the multiple small-scale feature maps are input into the preset convolutional neural network. Since the convolutional operations required for small-scale feature maps are fewer, this application can reduce the convolutional operations in the preset convolutional neural network, thereby reducing the computational amount and storage space required by the preset convolutional neural network. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] To more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0018] Figure 1 is a schematic flowchart of the implementation of the signal processing method provided in Embodiment 1 of the present application;

[0019] Figure 2 is a schematic flowchart of the implementation of the signal processing method provided in Embodiment 2 of the present application;

[0020] Figure 3 is a processing example diagram of the signal to be processed;

[0021] Figure 4 is a schematic flowchart of the implementation of the signal processing method provided in Embodiment 3 of the present application;

[0022] Figure 5 is another processing example diagram of the signal to be processed;

[0023] Figure 6 is a schematic structural diagram of the signal processing device provided in the fourth embodiment of the present application;

[0024] Figure 7 is a schematic structural diagram of the terminal device provided in the fifth embodiment of the present application. Detailed implementation manners

[0025] In the following description, for the purpose of illustration rather than limitation, specific details such as specific system architectures and technologies are presented in order to provide a thorough understanding of the embodiments of the present application. However, those skilled in the art should understand that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary details.

[0026] It should be understood that when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, wholes, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0027] It should also be understood that the terms used in this specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in this specification of the present application and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to include the plural forms.

[0028] In specific implementations, the terminal devices described in the embodiments of the present application include, but are not limited to, other portable devices such as mobile phones, laptop computers, or tablet computers having a touch-sensitive surface (e.g., a touch screen display and / or a touchpad). It should also be understood that in some embodiments, the device is not a portable communication device, but a desktop computer having a touch-sensitive surface (e.g., a touch screen display and / or a touchpad).

[0029] In the following discussion, terminal devices including a display and a touch-sensitive surface are described. However, it should be understood that the terminal device may include one or more other physical user interface devices such as a physical keyboard, a mouse, and / or a joystick.

[0030] The terminal device supports various application programs, such as one or more of the following: drawing application programs, presentation application programs, word processing application programs, website creation application programs, disc burning application programs, spreadsheet application programs, game application programs, telephone application programs, video conferencing application programs, email application programs, instant messaging application programs, exercise support application programs, photo management application programs, digital camera application programs, digital video camera application programs, web browsing application programs, digital music player application programs, and / or digital video player application programs.

[0031] Various application programs that can be executed on the terminal device can use at least one common physical user interface device such as a touch-sensitive surface. One or more functions of the touch-sensitive surface and the corresponding information displayed on the terminal can be adjusted and / or changed between application programs and / or within the corresponding application programs. In this way, the common physical architecture of the terminal (e.g., the touch-sensitive surface) can support various application programs with a user interface that is intuitive and transparent to the user.

[0032] It should be understood that the magnitudes of the sequence numbers of the steps in this embodiment do not imply the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0033] To illustrate the technical solutions described in the present application, the following will be described through specific embodiments.

[0034] See Figure 1 , which is a schematic implementation flowchart of the signal processing method provided in the first embodiment of the present application. The signal processing method is applied to a terminal device. As Figure 1 shown, the signal processing method may include the following steps:

[0035] Step 101, obtain the signal to be processed.

[0036] Among them, the signal to be processed may refer to a one-dimensional signal, a two-dimensional image, a three-dimensional image, etc., which are not limited herein.

[0037] The above-mentioned signal to be processed may be obtained by the terminal device from its own memory, or may be collected in real time by a signal acquisition device built in the terminal device, or may be obtained by the terminal device from other devices, which are not limited herein.

[0038] For example, when the terminal device receives a signal acquisition instruction, it can obtain the signal to be processed from its own memory, or can collect the signal to be processed in real time through a built-in signal acquisition device. The terminal device can also send the signal acquisition instruction to other devices, and after receiving the signal acquisition instruction, the other devices send the signal to be processed to the terminal device. The above-mentioned signal acquisition device can refer to a device capable of collecting signals. For example, when the signal to be processed is a two-dimensional image, the above-mentioned signal acquisition device can be a monocular camera, a binocular camera, etc.

[0039] Step 102: Decompose the signal to be processed into feature maps of N scales.

[0040] Wherein, N is an integer greater than 1, and the scale of the signal to be processed is greater than the maximum scale among the N scales.

[0041] It should be noted that the number of feature maps of each scale among the N scales can be one or at least two, and no limitation is made here. The number of channels of the signal to be processed can be different from the number of channels of the feature maps of the N scales. For example, the number of channels of the signal to be processed is 1, and the number of channels of the feature maps of the N scales is 3.

[0042] For example, when the signal to be processed is a two-dimensional image and the scale of the signal to be processed is 256*256, the terminal device can decompose the signal to be processed into feature maps of five scales, and the above five scales are 128*128, 64*64, 32*32, 16*16, and 8*8 respectively. Specifically, the signal to be processed can be decomposed into three feature maps of 128*128, three feature maps of 64*64, three feature maps of 32*32, three feature maps of 16*16, and four feature maps of 8*8.

[0043] Since the scale of the signal to be processed is greater than the maximum scale among the N scales, when the terminal device decomposes the signal to be processed into feature maps of N scales, it decomposes the large-scale signal to be processed into multiple small-scale feature maps. Inputting the multiple small-scale feature maps into the preset convolutional neural network can reduce the convolutional operations in the preset convolutional neural network, thereby reducing the computational amount and storage space required by the preset convolutional neural network, that is, reducing the computational amount and storage space occupied by the terminal device when running the preset convolutional neural network.

[0044] As an optional embodiment, the terminal device can perform wavelet decomposition on the signal to be processed to obtain feature maps of N scales.

[0045] Wavelet decomposition can be called wavelet transform. Wavelet transform is a time-scale analysis method for signals. It has the characteristic of multi-resolution and has the ability to characterize the local features of signals in both the time and frequency domains. It is a time-frequency localization analysis method with a fixed window size but a variable shape, and both the time window and the frequency window are variable. That is, wavelet transform has a high frequency resolution and time resolution in the low-frequency part, and a high time resolution and a low frequency resolution in the high-frequency part, which is more suitable for detecting transient anomalies in signals and displaying their components.

[0046] Wavelet decomposition has the characteristic of multi-resolution and can perform multi-scale filtering decomposition on the signal to be processed, forming feature maps with different frequency characteristics. The above-mentioned feature maps with different frequency characteristics are the feature maps of N scales.

[0047] Step 103: Input the feature maps of N scales into a preset convolutional neural network to obtain a target feature map.

[0048] Among them, the scale of the target feature map is smaller than the smallest scale among the N scales.

[0049] After the terminal device inputs the feature maps of N scales into the preset convolutional neural network, the preset convolutional neural network can process the feature maps of N scales, so as to output a target feature map that meets the requirements. For example, it can be required that the scale of the target feature map is 4*4.

[0050] The above-mentioned preset convolutional neural network can refer to any convolutional neural network that can output a target feature map.

[0051] In this embodiment, after the terminal device obtains the target feature map, it can determine the processing result of the signal to be processed according to the target feature map.

[0052] Among them, the processing result of the above-mentioned signal to be processed can refer to the classification result of the signal to be processed or the segmentation result of the signal to be processed, which is not limited here.

[0053] As an optional embodiment, the terminal device can input the target feature map into a fully connected layer or a hidden layer to obtain the processing result of the signal to be processed.

[0054] After the terminal device inputs the target feature map into the fully connected layer or the hidden layer, the fully connected layer or the hidden layer processes the target feature map and outputs the processing result of the signal to be processed.

[0055] After the terminal device obtains the processing result of the signal to be processed, it can display the above-mentioned processing result on its own display screen, or send the above-mentioned processing result to other devices so that users can obtain the above-mentioned processing result.

[0056] In the embodiments of the present application, by first decomposing a large-scale signal to be processed into multiple small-scale feature maps, and then inputting the multiple small-scale feature maps into a preset convolutional neural network, since the convolutional operations required for the small-scale feature maps are fewer, the convolutional operations in the preset convolutional neural network can be reduced. That is, compared with the prior art, under the premise of achieving the same network depth, the present application can effectively reduce the computational amount and storage space required by the preset convolutional neural network. That is to say, under the condition of the same computing power and storage resources, the present application can deepen the network structure and obtain better signal processing results.

[0057] See Figure 2 , which is a schematic flowchart of the implementation of the signal processing method provided in the second embodiment of the present application. This signal processing method is applied to a terminal device. As Figure 2 shown, this signal processing method may include the following steps:

[0058] Step 201, obtain a signal to be processed.

[0059] This step is the same as step 101. For specific details, please refer to the relevant description of step 101, which will not be elaborated here.

[0060] Step 202, decompose the signal to be processed into feature maps of N scales.

[0061] This step is the same as step 102. For specific details, please refer to the relevant description of step 102, which will not be elaborated here.

[0062] In one embodiment, the preset convolutional neural network for processing the feature maps of N scales may include network blocks corresponding to each of the N scales. The N scales are arranged in descending order. The terminal device uses the network blocks to process the feature maps of N scales, and can obtain target feature maps.

[0063] Among them, the above network blocks sequentially include a first convolutional layer and a pooling layer. The first convolutional layer can first perform convolutional calculations on the feature maps input to the network block to extract features and obtain the convolutional feature maps. The pooling layer performs dimensionality reduction processing on the convolutional feature maps to obtain the first feature maps. The pooling layer can reduce the dimension of the feature maps, reduce overfitting, and at the same time improve the fault tolerance of the preset convolutional neural network.

[0064] Step 203, input the feature map of the largest scale into the network block corresponding to the largest scale to obtain the first feature map corresponding to the largest scale.

[0065] Among them, the scale of the first feature map corresponding to the largest scale is the same as the next scale of the largest scale. Since the N scales are arranged in descending order, the next scale of the largest scale can be understood as the scale that is second only to the largest scale among the N scales or the scale arranged after the largest scale.

[0066] Step 204: Take the next scale of the largest scale as the target scale.

[0067] Step 205: Input the feature map of the target scale and the first feature map corresponding to the previous scale of the target scale into the network block corresponding to the target scale, and obtain the first feature map corresponding to the target scale.

[0068] Among them, the previous scale of the target scale may refer to the scale that is one position before the target scale among the N scales. For example, if the signal to be processed is a two-dimensional image and the signal to be processed is decomposed into feature maps of five scales, the five scales are 128*128, 64*64, 32*32, 16*16, and 8*8 respectively. The above five scales can be arranged in powers of 2, and the arrangement order is 128*128, 64*64, 32*32, 16*16, 8*8. Then the largest scale is 128*128, the next scale of the largest scale is 64*64, the previous scale of 64*64 is 128*128, the previous scale of 32*32 is 64*64, the previous scale of 16*16 is 32*32, and the previous scale of 8*8 is 16*16.

[0069] Step 206: Detect whether the target scale is the smallest scale among the N scales.

[0070] Step 207: Take the next scale of the target scale as the target scale.

[0071] Through the above steps 205 to 207, the terminal device can traverse the N scales, so as to obtain the first feature map corresponding to the smallest scale.

[0072] During the process of the terminal device traversing the N scales, gradually adding feature maps with different frequency characteristics to the network block can reduce the number of convolution kernels used by the preset convolutional neural network, thereby reducing the computational amount and storage space required by the preset convolutional neural network.

[0073] As an optional embodiment, the above preset convolutional neural network further includes a second convolutional layer corresponding to the target scale. Before inputting the feature map of the target scale and the first feature map corresponding to the previous scale of the target scale into the network block corresponding to the target scale, it further includes:

[0074] Input the feature map of the target scale and the first feature map corresponding to the previous scale of the target scale into the second convolutional layer corresponding to the target scale, and obtain the second feature map corresponding to the target scale;

[0075] The above inputting the feature map of the target scale and the first feature map corresponding to the previous scale of the target scale into the network block corresponding to the target scale and obtaining the first feature map corresponding to the target scale includes:

[0076] Input the second feature map corresponding to the target scale into the network block corresponding to the target scale to obtain the first feature map corresponding to the target scale.

[0077] Among them, the above-mentioned second convolutional layer is used to extract features.

[0078] Before inputting the feature map of the target scale and the first feature map corresponding to the previous scale of the target scale into the network block corresponding to the target scale, input them into the second convolutional layer first, which can further extract features and improve the accuracy of the processing result of the signal to be processed.

[0079] As Figure 3 shown is the processing example diagram of the signal to be processed. Figure 3 In Figure 3 the signal to be processed is a two-dimensional image, the scale of the signal to be processed is 256*256, and the number of channels is 1. Perform wavelet decomposition on the signal to be processed to decompose the signal to be processed into feature maps of five scales. The number of channels of the above five-scale feature maps is 3. The above five scales are 128*128, 64*64, 32*32, 16*16, and 8*8 respectively. Input the first feature map output by the network block corresponding to 8*8 into the fully connected layer, and the processing result of the signal to be processed can be obtained.

[0080] Step 208, determine the first feature map corresponding to the minimum scale as the target feature map.

[0081] Among them, the output of the preset convolutional neural network is the first feature map corresponding to the minimum scale, and this first feature map corresponding to the minimum scale is the target feature map.

[0082] In the embodiments of the present application, by gradually adding feature maps with different frequency characteristics to the preset convolutional neural network, the number of convolutional kernels used by the preset convolutional neural network can be reduced, thereby reducing the computational amount and storage space of the preset convolutional neural network.

[0083] See Figure 4 , which is the implementation process schematic diagram of the signal processing method provided in Embodiment 3 of the present application. This signal processing method is applied to a terminal device. As Figure 4 shown, this signal processing method may include the following steps:

[0084] Step 401, obtain the signal to be processed.

[0085] This step is the same as step 101. For specific details, please refer to the relevant description of step 101 and will not be repeated here.

[0086] Step 402, decompose the signal to be processed into feature maps of N scales.

[0087] This step is the same as step 102. For specific details, please refer to the relevant description of step 102 and will not be repeated here.

[0088] In one embodiment, the preset convolutional neural network for processing feature maps of N scales may include target convolutional neural networks corresponding to each of the N scales. Taking the j-th scale among the N scales as an example, where the j-th scale is any one of the above N scales, the feature map of the j-th scale can be processed through the target convolutional neural network corresponding to the j-th scale to output a third feature map corresponding to the j-th scale, and the scale of this third feature map is the same as the scale of the target feature map.

[0089] For example, if the signal to be processed is a two-dimensional image and the signal to be processed is decomposed into feature maps of five scales, the above five scales are 128*128, 64*64, 32*32, 16*16, and 8*8 respectively, and the scale of the target feature map is 4*4. Each of the above five scales corresponds to a target convolutional neural network, that is, the preset convolutional neural network includes five target convolutional neural networks, and the above five target convolutional neural networks all output third feature maps with a scale of 4*4.

[0090] Step 403: Input the feature maps of N scales into their respective corresponding target convolutional neural networks to obtain third feature maps corresponding to each of the N scales.

[0091] Inputting the feature maps of N scales into their respective corresponding target convolutional neural networks can be understood as inputting the feature maps of N scales into the target convolutional neural network in parallel.

[0092] Since the feature map of each scale contains information of a certain frequency, the target convolutional neural network corresponding to each scale can use fewer convolutional kernels to complete the convolution operation of the feature map of each scale, reducing the computational amount and storage space required by the preset convolutional neural network.

[0093] Step 404: Determine the third feature maps corresponding to each of the N scales as the target feature map.

[0094] After the terminal device obtains the third feature maps corresponding to each of the N scales through the N target convolutional neural networks, it can determine the set of the third feature maps corresponding to each of the N scales as the target feature map.

[0095] For example, when the signal to be processed is a two-dimensional image, the signal to be processed is decomposed into feature maps of five scales. The above five scales are 128*128, 64*64, 32*32, 16*16, and 8*8 respectively. The scale of the target feature map is 4*4. Each of the above five scales corresponds to a target convolutional neural network, that is, the preset convolutional neural network includes five target convolutional neural networks. Input the feature map with a scale of 128*128 into the target convolutional neural network corresponding to 128*128, and a third feature map with a scale of 4*4 can be output; input the feature map with a scale of 64*64 into the target convolutional neural network corresponding to 64*64, and a third feature map with a scale of 4*4 can be output; input the feature map with a scale of 32*32 into the target convolutional neural network corresponding to 32*32, and a third feature map with a scale of 4*4 can be output; input the feature map with a scale of 16*16 into the target convolutional neural network corresponding to 16*16, and a third feature map with a scale of 4*4 can be output; input the feature map with a scale of 8*8 into the target convolutional neural network corresponding to 8*8, and a third feature map with a scale of 4*4 can be output. The third feature maps with a scale of 4*4 output by the above five target convolutional neural networks are all target feature maps.

[0096] As Figure 5 shown is another processing example diagram of the signal to be processed. Figure 5 In Figure 5 it, the signal to be processed is a two-dimensional image, the scale of the signal to be processed is 256*256, and the number of channels is 1.

[0097] In the embodiment of the present application, by parallelly inputting feature maps of N scales into the target convolutional neural network, each target convolutional neural network can use fewer convolutional kernels to complete the convolution operation of the feature map, reducing the calculation amount and storage space required by the preset convolutional neural network.

[0098] It should be noted that in the examples of the above three method embodiments, the unit of the scale can be pixels.

[0099] Refer to Figure 6 , which is a schematic structural diagram of the signal processing device provided in the fourth embodiment of the present application. For the convenience of description, only the part related to the embodiment of the present application is shown.

[0100] The above signal processing device includes:

[0101] A signal acquisition module 61, configured to acquire a signal to be processed;

[0102] A signal decomposition module 62, configured to decompose the signal to be processed into feature maps of N scales, where N is an integer greater than 1, and the scale of the signal to be processed is greater than the maximum scale among the N scales;

[0103] A feature map input module 63, configured to input the feature maps of N scales into a preset convolutional neural network to obtain a target feature map, and the scale of the target feature map is smaller than the minimum scale among the N scales.

[0104] As an optional embodiment, the above-mentioned signal decomposition module 62 is specifically configured to:

[0105] Perform wavelet decomposition on the signal to be processed to obtain feature maps of N scales.

[0106] As an optional embodiment, the above-mentioned preset convolutional neural network includes network blocks corresponding to each of the N scales. The network block includes a first convolutional layer and a pooling layer. The N scales are arranged in descending order, and the N scales include the maximum scale and non-maximum scales. The above-mentioned feature map input module 63 includes:

[0107] A first input unit, configured to input the feature map of the maximum scale into the network block corresponding to the maximum scale to obtain a first feature map corresponding to the maximum scale, and the scale of the first feature map corresponding to the maximum scale is the same as the next scale of the maximum scale;

[0108] A first determination unit, configured to use the next scale of the maximum scale as the target scale;

[0109] A second input unit, configured to input the feature map of the target scale and the first feature map corresponding to the previous scale of the target scale into the network block corresponding to the target scale to obtain a first feature map corresponding to the target scale;

[0110] A scale detection unit, configured to detect whether the target scale is the minimum scale among the N scales;

[0111] A second determination unit, configured to, if the target scale is not the minimum scale, use the next scale of the target scale as the target scale and return to execute the second input unit;

[0112] A third determination unit, configured to, if the target scale is the minimum scale, determine the first feature map corresponding to the minimum scale as the target feature map.

[0113] As an optional embodiment, the preset convolutional neural network further includes a second convolutional layer corresponding to the target scale; the above-mentioned feature map input module 63 further includes:

[0114] A third input unit, configured to input a feature map of a target scale and a first feature map corresponding to the previous scale of the target scale into a second convolutional layer corresponding to the target scale, to obtain a second feature map corresponding to the target scale;

[0115] The above-mentioned second input unit is specifically configured to input the second feature map corresponding to the target scale into a network block corresponding to the target scale, to obtain a first feature map corresponding to the target scale.

[0116] As an optional embodiment, the preset convolutional neural network includes target convolutional neural networks corresponding to N scales respectively, and the above-mentioned feature map input module 63 includes:

[0117] A fourth input unit, configured to input the feature maps of N scales into the corresponding target convolutional neural networks respectively, to obtain third feature maps corresponding to N scales respectively, and the scale of the third feature map is the same as the scale of the target feature map;

[0118] A fourth determination unit, configured to determine the third feature maps corresponding to the feature maps of N scales respectively as the target feature map.

[0119] As an optional embodiment, the above-mentioned signal processing device further includes:

[0120] A result determination module, configured to determine a processing result of the signal to be processed according to the target feature map.

[0121] As an optional embodiment, the above-mentioned result determination module is specifically configured to:

[0122] Input the target feature map into a fully connected layer or a hidden layer, to obtain a processing result of the signal to be processed.

[0123] The signal processing device provided by the embodiments of the present application can be applied to the foregoing method embodiments. For details, please refer to the description of the foregoing method embodiments, which will not be elaborated here.

[0124] Figure 7 It is a schematic structural diagram of a terminal device provided in Embodiment 5 of the present application. As Figure 7 shown, the terminal device 7 of this embodiment includes: one or more processors 70 (only one is shown in the figure), a memory 71, and a computer program 72 stored in the memory 71 and executable on the processor 70. When the processor 70 executes the computer program 72, the steps in the foregoing various signal processing method embodiments are implemented

[0125] The terminal device 7 may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The terminal device 7 may include, but is not limited to, a processor 70 and a memory 71. Those skilled in the art can understand, Figure 7This is only an example of the terminal device 7, which does not constitute a limitation on the terminal device 7. It may include more or fewer components than those shown in the figure, or combine certain components, or different components. For example, the terminal device 7 may also include input / output devices, network access devices, buses, etc.

[0126] The so-called processor 70 may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor, or the processor may also be any conventional processor, etc.

[0127] The memory 71 may be an internal storage unit of the terminal device 7, such as the hard disk or memory of the terminal device 7. The memory 71 may also be an external storage device of the terminal device 7, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the terminal device 7. Further, the memory 71 may also include both the internal storage unit and the external storage device of the terminal device 7. The memory 71 is used to store the computer program and other programs and data required by the terminal device 7. The memory 71 may also be used to temporarily store data that has been output or is to be output.

[0128] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above functional units and modules is used as an example. In actual applications, the above functions can be assigned to different functional units and modules according to needs, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiments can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated unit can be implemented in the form of hardware or in the form of a software functional unit. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other and do not limit the protection scope of this application. The specific working processes of the units and modules in the above system can refer to the corresponding processes in the foregoing method embodiments and will not be described herein again.

[0129] In the above embodiments, the descriptions of the various embodiments have their own emphases. For the parts not detailed or recorded in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0130] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or by a combination of computer software and electronic hardware. Whether these functions are executed in the form of hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0131] In the embodiments provided in this application, it should be understood that the disclosed device / terminal device and method can be implemented in other ways. For example, the device / terminal device embodiments described above are only illustrative. For example, the division of the modules or units is only a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical or other forms.

[0132] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0133] In addition, in each embodiment of the present application, the functional units can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0134] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, to implement all or part of the processes in the above-mentioned embodiment methods of the present application, it can also be completed by a computer program instructing the relevant hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or some intermediate forms, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0135] To implement all or part of the processes in the above-mentioned embodiment methods of the present application, it can also be completed by a computer program product. When the computer program product runs on a terminal device, the terminal device can execute to implement the steps in the above-mentioned various method embodiments.

[0136] The above-mentioned embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A signal processing method, characterized in that, The signal processing method includes: Obtain a signal to be processed; Decompose the signal to be processed into feature maps of N scales, where N is an integer greater than 1, and the scale of the signal to be processed is greater than the maximum scale among the N scales; Input the feature maps of the N scales into a preset convolutional neural network to obtain a target feature map, and the scale of the target feature map is smaller than the minimum scale among the N scales; The preset convolutional neural network includes network blocks corresponding to each of the N scales. The network blocks sequentially include a first convolutional layer and a pooling layer. The N scales are arranged in descending order. The step of inputting the feature maps of the N scales into the preset convolutional neural network to obtain a target feature map includes: Input the feature map of the largest scale among the N scales into the network block corresponding to the largest scale to obtain a first feature map corresponding to the largest scale, and the scale of the first feature map corresponding to the largest scale is the same as the next scale of the largest scale; Take the next scale of the largest scale as the target scale; Input the feature map of the target scale and the first feature map corresponding to the previous scale of the target scale into the network block corresponding to the target scale to obtain a first feature map corresponding to the target scale; Detect whether the target scale is the smallest scale among the N scales; If the target scale is not the smallest scale, take the next scale of the target scale as the target scale, and return to execute the step of inputting the feature map of the target scale and the first feature map corresponding to the previous scale of the target scale into the network block corresponding to the target scale to obtain a first feature map corresponding to the target scale; If the target scale is the smallest scale, determine the first feature map corresponding to the smallest scale as the target feature map.

2. The signal processing method according to claim 1, wherein The step of decomposing the signal to be processed into feature maps of N scales includes: Perform wavelet decomposition on the signal to be processed to obtain the feature maps of the N scales.

3. The signal processing method according to claim 1, wherein The preset convolutional neural network further includes a second convolutional layer corresponding to the target scale. Before inputting the feature map of the target scale and the first feature map corresponding to the previous scale of the target scale into the network block corresponding to the target scale, it further includes: Input the feature map of the target scale and the first feature map corresponding to the previous scale of the target scale into the second convolutional layer corresponding to the target scale to obtain a second feature map corresponding to the target scale; The step of inputting the feature map of the target scale and the first feature map corresponding to the previous scale of the target scale into the network block corresponding to the target scale to obtain a first feature map corresponding to the target scale includes: Input the second feature map corresponding to the target scale into the network block corresponding to the target scale to obtain a first feature map corresponding to the target scale.

4. The signal processing method according to claim 1, wherein, The preset convolutional neural network includes target convolutional neural networks corresponding to each of the N scales. The step of inputting the feature maps of the N scales into the preset convolutional neural network to obtain a target feature map includes: Input the feature maps of the N scales into their respective corresponding target convolutional neural networks to obtain third feature maps corresponding to the N scales respectively, where the scale of the third feature map is the same as the scale of the target feature map; Determine the third feature maps corresponding to the feature maps of the N scales respectively as the target feature map.

5. The signal processing method according to any one of claims 1 to 4, characterized in that, After obtaining the target feature map, it further includes: Determine the processing result of the signal to be processed according to the target feature map.

6. The signal processing method according to claim 5, wherein, The determining the processing result of the signal to be processed according to the target feature map includes: Input the target feature map into a fully connected layer or a hidden layer to obtain the processing result of the signal to be processed.

7. A signal processing device, characterized in that, The signal processing device includes: A signal acquisition module, configured to acquire a signal to be processed; A signal decomposition module, configured to decompose the signal to be processed into feature maps of N scales, where N is an integer greater than 1, and the scale of the signal to be processed is greater than the maximum scale among the N scales; A feature map input module, configured to input the feature maps of the N scales into a preset convolutional neural network to obtain a target feature map, where the scale of the target feature map is smaller than the minimum scale among the N scales; The preset convolutional neural network includes network blocks corresponding to the N scales respectively, the network blocks sequentially include a first convolutional layer and a pooling layer, the N scales are arranged in descending order, and the feature map input module includes: A first input unit, configured to input the feature map with the largest scale among the N scales into the network block corresponding to the largest scale to obtain a first feature map corresponding to the largest scale, where the scale of the first feature map corresponding to the largest scale is the same as the next scale of the largest scale; A first determination unit, configured to use the next scale of the largest scale as the target scale; A second input unit, configured to input the feature map of the target scale and the first feature map corresponding to the previous scale of the target scale into the network block corresponding to the target scale to obtain a first feature map corresponding to the target scale; A scale detection unit, configured to detect whether the target scale is the smallest scale among the N scales; A second determination unit, configured to, if the target scale is not the smallest scale, use the next scale of the target scale as the target scale and return to execute the step of inputting the feature map of the target scale and the first feature map corresponding to the previous scale of the target scale into the network block corresponding to the target scale to obtain a first feature map corresponding to the target scale; A third determination unit, configured to, if the target scale is the smallest scale, determine the first feature map corresponding to the smallest scale as the target feature map.

8. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the signal processing method according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the signal processing method according to any one of claims 1 to 6.

Images