Interest analysis product recommendation method and system for e-commerce platforms based on artificial intelligence

Abstract

The present invention discloses an interest analysis-based commodity recommendation method and system for an e-commerce platform based on artificial intelligence, and relates to the field of data processing technology. The method comprises: extracting a user voice signal and a video signal containing a user's face; intercepting an arbitrary video signal as a video to be detected, and performing identification verification on the user's identity. If the verification result is that the user's identity is consistent with the user account, the user's voice signal is identified to obtain a voice recognition result; performing concentration detection on each frame of the video signal to obtain multiple concentration detection results; determining and recording commodities with high user interest based on the voice recognition result and the multiple concentration detection results, and extracting related commodities in the e-commerce platform system to form a commodity recommendation set, and recommending it to the corresponding user. The present invention combines a variety of models to achieve accurate user identity recognition verification, voice recognition and concentration detection, thereby achieving accurate and effective commodity recommendations.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to an interest analysis-based product recommendation method and system for an e-commerce platform based on artificial intelligence. Background Art

[0002] With the development of internet technology and mobile payments, e-commerce platforms have become a mainstream way of shopping in modern society. For users, e-commerce platforms offer a wider variety of products and specifications, meeting their growing personalized needs. For merchants, e-commerce platforms not only reduce operating costs but also increase exposure and sales.

[0003] However, due to the large variety of products on e-commerce platforms, users often struggle to efficiently select their desired products, significantly increasing their time spent selecting products. With the development of artificial intelligence (AI), it is now possible to fully analyze user interests and preferences, providing direct support for targeted product recommendations. Therefore, proposing an AI-based interest-based product recommendation method and system for e-commerce platforms is of great value and significance. Summary of the Invention

[0004] In order to overcome the above problems or at least partially solve the above problems, the present invention provides an interest analysis-based product recommendation method and system for an e-commerce platform based on artificial intelligence, which combines a multi-dimensional image coding matching identity verification model based on video frame optimization, an end-to-end speech recognition model based on speech denoising and the joint application of multiple recognition modules, and an adaptive consumption-based concentration detection model based on classification confidence analysis to achieve accurate user identity recognition verification, speech recognition and concentration detection, thereby achieving accurate and effective product recommendations.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] In a first aspect, the present invention provides an interest analysis-based product recommendation method for an e-commerce platform based on artificial intelligence, comprising the following steps:

[0007] After the user logs in to the e-commerce platform system using the corresponding user account and password, the user's voice signal and the video signal containing the user's face are extracted based on the audio and video extraction device;

[0008] Intercept any video signal as the video to be detected, use the multi-dimensional image coding matching identity verification model based on video frame optimization to identify and verify the user's identity and generate a verification result;

[0009] If the verification result shows that the user identity is consistent with the user account, the end-to-end speech recognition model based on speech denoising and multi-recognition module is used to recognize the user's speech signal to obtain a speech recognition result;

[0010] Using an adaptive consumption-based focus detection model based on classification confidence analysis, the focus detection is performed on each frame of the video signal to obtain multiple focus detection results;

[0011] Based on the speech recognition results and multiple concentration detection results, the products that the user is most interested in are determined and recorded, and relevant products are extracted from the e-commerce platform system to form a product recommendation set, which is then recommended to the corresponding user.

[0012] First, the present invention proposes a multi-dimensional image encoding matching-based identity verification model based on video frame optimization to verify user identity. This model selects high-quality video frames and then performs multi-dimensional encoding matching between these high-quality video frames and template facial images, thereby achieving more accurate identity verification. Second, the present invention proposes an end-to-end speech recognition model based on the combined application of speech denoising and multiple recognition modules to recognize speech signals. This model can jointly apply a speech denoising network module, multiple speech recognition network modules, and a speech recognition result comparison module to achieve end-to-end accurate speech recognition. Finally, the present invention also proposes an adaptive consumption-based concentration detection model based on classification confidence analysis to perform concentration detection on each frame in the video signal. When using a simple concentration detection model, if a concentration detection result with a high classification confidence is already obtained, there is no need to use a more complex concentration detection model; otherwise, a complex concentration detection model is used to complete the concentration detection. Using this concentration detection method not only ensures the accuracy of concentration detection but also reduces computing resource consumption.

[0013] Based on the first aspect, further, the method for identifying and verifying a user's identity using the multi-dimensional image coding matching identity verification model based on video frame optimization includes the following steps:

[0014] Use the peak signal-to-noise ratio detection model to detect each frame image in the video to be detected, so as to select multiple high-quality frame images;

[0015] Using a high-dimensional autoencoder to perform high-dimensional encoding on each high-quality frame image and a preset template face image, and calculating the similarity between each high-quality frame image and the template face image to obtain multiple high-dimensional similarity results;

[0016] Using a low-dimensional autoencoder to perform low-dimensional encoding on each high-quality frame image and a preset template face image, and calculating the similarity between each high-quality frame image and the template face image to obtain multiple low-dimensional similarity results;

[0017] If all high-dimensional similarity results and low-dimensional similarity results are higher than the preset similarity threshold, the user identity is deemed to be consistent with the user account.

[0018] Based on the first aspect, further, the method for recognizing a user's voice signal using an end-to-end voice recognition model based on voice denoising and a joint application of multiple recognition modules includes the following steps:

[0019] Multiple speech recognition network modules using different speech recognition algorithms are connected in parallel, and a speech denoising network module is connected to the front end of the multiple speech recognition network modules, and a speech recognition result comparison module is connected to the back end of the multiple speech recognition network modules to form an end-to-end speech recognition model.

[0020] User voice signals are recognized through an end-to-end speech recognition model.

[0021] Based on the first aspect, further, the method for recognizing a user voice signal using an end-to-end voice recognition model includes the following steps:

[0022] Input the user's voice signal into the voice denoising network module for denoising to obtain a denoised voice signal;

[0023] Inputting the denoised speech signal into multiple speech recognition network modules for speech recognition to obtain multiple recognition results;

[0024] The multiple recognition results are input into the speech recognition result comparison module for comparison and analysis. If the multiple recognition results are consistent, the final speech recognition result is output.

[0025] Based on the first aspect, further, the method for performing concentration detection on each frame of a video signal using the adaptive consumption-based concentration detection model based on classification confidence analysis includes the following steps:

[0026] Select several facial images with high concentration as positive samples, and select several images of people with low concentration as negative samples;

[0027] Select some positive samples and some negative samples to train the SVM model to obtain a concentration detection model based on the SVM model;

[0028] Use all positive samples and all negative samples to train the convolutional neural network to obtain a concentration detection model based on the convolutional neural network;

[0029] For any frame of image, multi-scale reconstruction is performed to obtain images of multiple scales;

[0030] Use the concentration detection model based on the SVM model to detect images of multiple scales to obtain multiple detection results;

[0031] If multiple detection results are consistent, the final concentration detection result is output; otherwise, the image is detected using a concentration detection model based on a convolutional neural network to obtain the final concentration detection result.

[0032] In a second aspect, the present invention provides an interest analysis-based product recommendation system for an e-commerce platform based on artificial intelligence, comprising a signal extraction module, an identity verification module, a speech recognition module, a concentration detection module, and a product recommendation module, wherein:

[0033] The signal extraction module is used to extract the user's voice signal and the video signal containing the user's face based on the audio and video extraction device after the user logs in to the e-commerce platform system using the corresponding user account and password;

[0034] The identity verification module is used to intercept any video signal as the video to be detected, use the multi-dimensional image coding matching identity verification model based on video frame optimization to identify and verify the user's identity and generate a verification result;

[0035] A speech recognition module is used to recognize the user's speech signal using an end-to-end speech recognition model based on speech denoising and a combination of multiple recognition modules to obtain a speech recognition result if the verification result shows that the user's identity is consistent with the user account;

[0036] A concentration detection module is used to perform concentration detection on each frame of the video signal using an adaptive consumption-based concentration detection model based on classification confidence analysis to obtain multiple concentration detection results;

[0037] The product recommendation module is used to determine and record products that users are highly interested in based on voice recognition results and multiple concentration detection results, extract related products from the e-commerce platform system, form a product recommendation set, and recommend it to the corresponding users.

[0038] This system achieves accurate user identity verification, voice recognition, and concentration detection through the combination of multiple modules such as signal extraction module, identity verification module, voice recognition module, concentration detection module, and product recommendation module, thereby achieving accurate and effective product recommendations. First, this system utilizes a multi-dimensional image coding matching identity verification model based on video frame optimization to verify user identity; this model selects high-quality frame images of the video, and on this basis, performs multi-dimensional coding matching on the high-quality frame images of the video and the template face image, thereby achieving more accurate identity verification. Secondly, this system utilizes an end-to-end speech recognition model based on the joint application of speech denoising and multiple recognition modules to recognize speech signals; this model can jointly apply a speech denoising network module, multiple speech recognition network modules, and a speech recognition result comparison module to achieve end-to-end accurate speech recognition. Finally, this system utilizes an adaptive, cost-effective focus detection model based on classification confidence analysis to perform focus detection on each frame of the video signal. If a simple focus detection model can produce a focus detection result with high classification confidence, there's no need to use a more complex focus detection model. Otherwise, a more complex focus detection model is used to complete the focus detection. This focus detection approach not only ensures focus detection accuracy but also reduces computational resource consumption.

[0039] In a third aspect, the present application provides an electronic device comprising a memory for storing one or more programs; a processor; and when the one or more programs are executed by the processor, the method of any one of the above-mentioned first aspects is implemented.

[0040] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in any one of the first aspects above.

[0041] The present invention has at least the following advantages or beneficial effects:

[0042] The present invention provides an interest analysis-based product recommendation method and system for an e-commerce platform based on artificial intelligence. By combining a multi-dimensional image coding matching identity verification model based on video frame optimization, an end-to-end speech recognition model based on speech denoising and the joint application of multiple recognition modules, and an adaptive consumption-based concentration detection model based on classification confidence analysis, accurate user identity recognition verification, speech recognition, and concentration detection are achieved, thereby realizing accurate and effective product recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0044] Figure 1 This is a flowchart of an interest analysis-based product recommendation method for an e-commerce platform based on artificial intelligence according to an embodiment of the present invention;

[0045] Figure 2 This is a flowchart of user identity verification in an interest analysis-based product recommendation method for an e-commerce platform based on artificial intelligence according to an embodiment of the present invention;

[0046] Figure 3 This is a flowchart of voice recognition in an interest analysis-based product recommendation method for an e-commerce platform based on artificial intelligence according to an embodiment of the present invention;

[0047] Figure 4 This is a functional block diagram of an interest analysis-based product recommendation system for an e-commerce platform based on artificial intelligence according to an embodiment of the present invention;

[0048] Figure 5 This is a structural block diagram of an electronic device provided by an embodiment of the present invention.

[0049] Explanation of the accompanying drawings: 100, signal extraction module; 200, identity verification module; 300, speech recognition module; 400, concentration detection module; 500, product recommendation module; 101, memory; 102, processor; 103, communication interface. DETAILED DESCRIPTION

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0051] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0052] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not require further definition or explanation in subsequent drawings.

[0053] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprises" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further limitations, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device that includes the element.

[0054] In the description of the embodiments of the present invention, "a plurality of" means at least two.

[0055] Example:

[0056] like Figure 1-Figure 3 As shown, in a first aspect, the present invention provides an interest analysis-based product recommendation method for an e-commerce platform based on artificial intelligence, comprising the following steps:

[0057] S1. After a user logs in to the e-commerce platform system using the corresponding user account and password, the user's voice signal and the video signal containing the user's face are extracted using an audio and video extraction device;

[0058] S2. Intercept any video signal as the video to be detected, use the multi-dimensional image coding matching identity verification model based on video frame optimization to identify and verify the user's identity, and generate a verification result;

[0059] Further, if Figure 2 Shown, including:

[0060] S21, using a peak signal-to-noise ratio detection model to detect each frame image in the video to be detected, so as to select multiple high-quality frame images;

[0061] S22, using a high-dimensional autoencoder to perform high-dimensional encoding on each high-quality frame image and a preset template face image, and calculating the similarity between each high-quality frame image and the template face image to obtain multiple high-dimensional similarity results;

[0062] S23, using a low-dimensional autoencoder to perform low-dimensional encoding on each high-quality frame image and a preset template face image, and calculating the similarity between each high-quality frame image and the template face image to obtain multiple low-dimensional similarity results;

[0063] S24. If all high-dimensional similarity results and low-dimensional similarity results are higher than a preset similarity threshold, it is determined that the user identity is consistent with the user account.

[0064] In some embodiments of the present invention, a portion of the video signal (3-5 seconds is sufficient) is captured as the video to be tested. A multi-dimensional image encoding matching-based identity verification model based on video frame optimization is used to verify the user's identity. If the user's identity matches the account's corresponding identity, the process proceeds to the next step.

[0065] The aforementioned multi-dimensional image encoding matching identity verification model based on video frame optimization specifically involves: using a peak signal-to-noise ratio (PSNR) detection model to detect each frame in the video to be detected, selecting several (usually 10) frames with high PSNR as high-quality frames. The first high-quality frame image and the template facial image (the template facial image corresponding to the account is already stored in the system) are each encoded in a high-dimensional manner using a high-dimensional autoencoder, and their similarity is calculated using Euclidean distance to obtain the high-dimensional encoding similarity. Then, the two are each encoded in a low-dimensional manner using a low-dimensional autoencoder, and their similarity is calculated using Euclidean distance to obtain the low-dimensional encoding similarity. If the encoding similarities across different dimensions are all high, the first high-quality frame image and the template facial image are considered to have a high similarity. The similarities between the remaining high-quality frames and the template facial image are calculated using the same method. If all high-quality frames and the template facial image have a high similarity, the user identity is considered to be consistent with the corresponding account identity. Note: High-dimensional autoencoders can produce encoding results with higher dimensions; low-dimensional autoencoders can produce encoding results with lower dimensions.

[0066] S3. If the verification result shows that the user identity is consistent with the user account, an end-to-end speech recognition model based on speech denoising and a joint application of multiple recognition modules is used to recognize the user's speech signal to obtain a speech recognition result;

[0067] Furthermore, it includes: connecting multiple speech recognition network modules using different speech recognition algorithms in parallel, and connecting a speech denoising network module to the front end of the multiple speech recognition network modules, and connecting a speech recognition result comparison module to the back end of the multiple speech recognition network modules to form an end-to-end speech recognition model; recognizing user voice signals through the end-to-end speech recognition model.

[0068] Further, if Figure 3 Shown, including:

[0069] S31, inputting the user voice signal into the voice denoising network module for denoising processing to obtain a denoised voice signal;

[0070] S32, inputting the denoised speech signal into a plurality of speech recognition network modules for speech recognition to obtain a plurality of recognition results;

[0071] S33: Input the multiple recognition results into the speech recognition result comparison module for comparison and analysis. If the multiple recognition results are consistent, the final speech recognition result is output.

[0072] In some embodiments of the present invention, an end-to-end speech recognition model based on speech denoising and the combined application of multiple recognition modules is used to recognize speech signals. If the user utters words such as "great" or "I like" expressing appreciation, satisfaction, or approval, the user is deemed to have a high level of interest in the product being viewed.

[0073] The above-mentioned end-to-end speech recognition model based on speech denoising and the joint application of multiple recognition modules specifically includes: connecting speech recognition network module A, speech recognition network module B, and speech recognition network module C in parallel (the speech recognition algorithms used by the several network modules are different), connecting a speech denoising network module at their front end, and connecting a speech recognition result comparison module at their back end. The above modules together constitute an end-to-end speech recognition model. First, the input speech signal passes through the speech denoising network module to obtain a denoised speech signal. Secondly, the denoised speech signal passes through the speech recognition network module A to obtain speech recognition result A; the denoised speech signal passes through the speech recognition network module B to obtain speech recognition result B; the denoised speech signal passes through the speech recognition network module C to obtain speech recognition result C. Finally, speech recognition result A, speech recognition result B, and speech recognition result C pass through the speech recognition result comparison module together. If the three speech recognition results are consistent, the final speech recognition result will be output.

[0074] S4. Using an adaptive consumption-based concentration detection model based on classification confidence analysis, perform concentration detection on each frame of the video signal to obtain multiple concentration detection results;

[0075] Furthermore, it includes: selecting several facial images with high concentration as positive samples, and selecting several images of people with low concentration as negative samples; selecting some positive samples and some negative samples to train the SVM model to obtain a concentration detection model based on the SVM model; using all positive samples and all negative samples to train the convolutional neural network to obtain a concentration detection model based on the convolutional neural network; performing multi-scale reconstruction on any frame image to obtain images of multiple scales; using the concentration detection model based on the SVM model to detect images of multiple scales to obtain multiple detection results; if the multiple detection results are consistent, outputting the final concentration detection result; otherwise, using the concentration detection model based on the convolutional neural network to detect the image to obtain the final concentration detection result.

[0076] In some embodiments of the present invention, for the vast majority of all frame images within a period of time (about 5 seconds), it can be detected that the user's concentration is high, and it is determined that the user has a high interest in the product being viewed at that time.

[0077] The above-mentioned adaptive consumption-based concentration detection model based on classification confidence analysis specifically includes: selecting a sufficient number of facial images with high concentration as positive samples, and selecting a sufficient number of facial images with low concentration as negative samples. Select some positive samples and some negative samples, train the SVM model, and obtain a concentration detection model based on SVM. Use all positive samples and all negative samples to train the convolutional neural network to obtain a concentration detection model based on the convolutional neural network. For a certain frame image, it is constructed into images of multiple scales, and the SVM-based concentration detection model is used to detect images of multiple scales separately. If the detection results of images of multiple scales are consistent (the concentration is both high or both low), the concentration detection result is directly obtained; if the detection results of images of multiple scales are inconsistent, the frame image is detected using the convolutional neural network-based concentration detection model (detection can be performed at the original scale) to obtain the final concentration detection result.

[0078] S5. Determine and record products that the user is most interested in based on the voice recognition results and multiple concentration detection results, extract related products from the e-commerce platform system, form a product recommendation set, and recommend it to the corresponding user.

[0079] In the above steps S3 and S4, if any one of the steps indicates that the user has a high interest in the product being viewed, the product being viewed is directly recorded, and products of the same category or other highly related categories are recommended to the user. If any one of the steps indicates that the user has a high interest in the product being viewed, the product being viewed is directly recorded, and products of the same category or other highly related categories are recommended to the user.

[0080] First, the present invention proposes a multi-dimensional image encoding matching-based identity verification model based on video frame optimization to verify user identity. This model selects high-quality video frames and then performs multi-dimensional encoding matching between these high-quality video frames and template facial images, thereby achieving more accurate identity verification. Second, the present invention proposes an end-to-end speech recognition model based on the combined application of speech denoising and multiple recognition modules to recognize speech signals. This model can jointly apply a speech denoising network module, multiple speech recognition network modules, and a speech recognition result comparison module to achieve end-to-end accurate speech recognition. Finally, the present invention also proposes an adaptive consumption-based concentration detection model based on classification confidence analysis to perform concentration detection on each frame in the video signal. When using a simple concentration detection model, if a concentration detection result with a high classification confidence is already obtained, there is no need to use a more complex concentration detection model; otherwise, a complex concentration detection model is used to complete the concentration detection. Using this concentration detection method not only ensures the accuracy of concentration detection but also reduces computing resource consumption.

[0081] like Figure 4 As shown, in a second aspect, an embodiment of the present invention provides an interest analysis-based product recommendation system for an e-commerce platform based on artificial intelligence, comprising a signal extraction module 100, an identity verification module 200, a speech recognition module 300, a concentration detection module 400, and a product recommendation module 500, wherein:

[0082] The signal extraction module 100 is used to extract the user's voice signal and the video signal containing the user's face based on the audio and video extraction device after the user logs in to the e-commerce platform system using the corresponding user account and password;

[0083] The identity verification module 200 is used to intercept any video signal as the video to be detected, use the multi-dimensional image coding matching identity verification model based on video frame optimization to identify and verify the user's identity and generate a verification result;

[0084] The speech recognition module 300 is configured to recognize the user's speech signal using an end-to-end speech recognition model based on speech denoising and a combination of multiple recognition modules to obtain a speech recognition result if the verification result shows that the user identity is consistent with the user account;

[0085] A concentration detection module 400 is configured to perform concentration detection on each frame of a video signal using an adaptive consumption-based concentration detection model based on classification confidence analysis to obtain a plurality of concentration detection results;

[0086] The product recommendation module 500 is used to determine and record products of high user interest based on the voice recognition results and multiple concentration detection results, and extract related products from the e-commerce platform system to form a product recommendation set and recommend it to the corresponding user.

[0087] This system achieves accurate user identity verification, voice recognition, and concentration detection through the combination of multiple modules such as the signal extraction module 100, the identity verification module 200, the speech recognition module 300, the concentration detection module 400, and the product recommendation module 500, thereby achieving accurate and effective product recommendations. First, this system utilizes a multi-dimensional image coding matching identity verification model based on video frame optimization to verify the user's identity; this model selects high-quality frame images of the video, and on this basis, performs multi-dimensional coding matching on the high-quality frame images of the video and the template face image, thereby achieving more accurate identity verification. Secondly, this system utilizes an end-to-end speech recognition model based on the joint application of speech denoising and multiple recognition modules to recognize speech signals; this model can jointly apply a speech denoising network module, multiple speech recognition network modules, and a speech recognition result comparison module to achieve end-to-end accurate speech recognition. Finally, this system utilizes an adaptive, cost-effective focus detection model based on classification confidence analysis to perform focus detection on each frame of the video signal. If a simple focus detection model can produce a focus detection result with high classification confidence, there's no need to use a more complex focus detection model. Otherwise, a more complex focus detection model is used to complete the focus detection. This focus detection approach not only ensures focus detection accuracy but also reduces computational resource consumption.

[0088] like Figure 5 As shown, in a third aspect, an embodiment of the present application provides an electronic device, which includes a memory 101 for storing one or more programs and a processor 102. When the one or more programs are executed by the processor 102, any method as described in the first aspect above is implemented.

[0089] The system also includes a communication interface 103. The memory 101, processor 102, and communication interface 103 are electrically connected to each other directly or indirectly to enable data transmission or interaction. For example, these components can be electrically connected to each other via one or more communication buses or signal lines. The memory 101 can be used to store software programs and modules, and the processor 102 executes the software programs and modules stored in the memory 101 to perform various functional applications and data processing. The communication interface 103 can be used to communicate signaling or data with other node devices.

[0090] Among them, the memory 101 can be, but is not limited to, random access memory (RAM), read only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0091] The processor 102 may be an integrated circuit chip with signal processing capabilities. The processor 102 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0092] In the embodiments provided in this application, it should be understood that the disclosed methods and systems can also be implemented in other ways. The method and system embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions, and operations of the methods and systems, methods, and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment, or a portion of code, and the module, program segment, or a portion of code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions.

[0093] In addition, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0094] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which implements a method as described in any one of the first aspects above when executed by the processor 102. If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or partly or partially embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0095] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

[0096] It will be apparent to those skilled in the art that the present application is not limited to the details of the exemplary embodiments described above and that the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the present application is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. An interest analysis-based product recommendation method for an e-commerce platform based on artificial intelligence, characterized in that: The following steps are involved: After the user logs in to the e-commerce platform system using the corresponding user account and password, the user's voice signal and the video signal containing the user's face are extracted based on the audio and video extraction device; Intercept any video signal as the video to be detected, use the multi-dimensional image coding matching identity verification model based on video frame optimization to identify and verify the user's identity and generate a verification result; If the verification result shows that the user identity is consistent with the user account, the end-to-end speech recognition model based on speech denoising and multi-recognition module is used to recognize the user's speech signal to obtain a speech recognition result; Using an adaptive consumption-based focus detection model based on classification confidence analysis, the focus detection is performed on each frame of the video signal to obtain multiple focus detection results; The method comprises the following steps: selecting a number of facial images with high concentration as positive samples, and selecting a number of images of people with low concentration as negative samples; selecting some positive samples and some negative samples to train the SVM model to obtain a concentration detection model based on the SVM model; using all positive samples and all negative samples to train the convolutional neural network to obtain a concentration detection model based on the convolutional neural network; performing multi-scale reconstruction on any frame image to obtain images of multiple scales; using the concentration detection model based on the SVM model to detect the images of multiple scales to obtain multiple detection results; if the multiple detection results are consistent, outputting the final concentration detection result; otherwise, using the concentration detection model based on the convolutional neural network to detect the image to obtain the final concentration detection result; Based on the speech recognition results and multiple concentration test results, products with high user interest are identified and recorded, and related products are extracted from the e-commerce platform system to form a product recommendation set, which is recommended to the corresponding user. If a result indicating that the user is highly interested in the product being viewed at that time is obtained in any of the above speech recognition and concentration test steps, the product being viewed by the user is directly recorded, and products in the same category and other highly related categories are recommended to the user for viewing; The method for identifying and verifying a user's identity using a multi-dimensional image coding matching identity verification model based on video frame optimization includes the following steps: Use the peak signal-to-noise ratio detection model to detect each frame image in the video to be detected, so as to select multiple high-quality frame images; Using a high-dimensional autoencoder to perform high-dimensional encoding on each high-quality frame image and a preset template face image, and calculating the similarity between each high-quality frame image and the template face image to obtain multiple high-dimensional similarity results; Using a low-dimensional autoencoder to perform low-dimensional encoding on each high-quality frame image and a preset template face image, and calculating the similarity between each high-quality frame image and the template face image to obtain multiple low-dimensional similarity results; If all high-dimensional similarity results and low-dimensional similarity results are higher than the preset similarity threshold, the user identity is deemed to be consistent with the user account; The method for recognizing a user's voice signal using an end-to-end voice recognition model based on voice denoising and a joint application of multiple recognition modules comprises the following steps: Multiple speech recognition network modules using different speech recognition algorithms are connected in parallel. A speech denoising network module is connected to the front end of each of the multiple speech recognition network modules, and a speech recognition result comparison module is connected to the back end of each of the multiple speech recognition network modules to form an end-to-end speech recognition model. User speech signals are recognized using the end-to-end speech recognition model. The method for recognizing a user's voice signal using an end-to-end voice recognition model comprises the following steps: The user voice signal is input into the voice denoising network module for denoising processing to obtain a denoised voice signal; the denoised voice signal is input into multiple voice recognition network modules for voice recognition to obtain multiple recognition results; the multiple recognition results are input into the voice recognition result comparison module for comparison and analysis. If the multiple recognition results are consistent, the final voice recognition result is output.

2. An interest analysis product recommendation system for e-commerce platforms based on artificial intelligence, characterized by: It includes a signal extraction module, an identity verification module, a speech recognition module, a concentration detection module, and a product recommendation module, among which: The signal extraction module is used to extract the user's voice signal and the video signal containing the user's face based on the audio and video extraction device after the user logs in to the e-commerce platform system using the corresponding user account and password; The identity verification module is used to intercept any video signal as the video to be detected, and use a multi-dimensional image coding matching identity verification model based on video frame optimization to identify and verify the user identity and generate a verification result; wherein: the use of the multi-dimensional image coding matching identity verification model based on video frame optimization to identify and verify the user identity includes: using a peak signal-to-noise ratio detection model to detect each frame image in the video to be detected to select multiple high-quality frame images; using a high-dimensional autoencoder to perform high-dimensional encoding on each high-quality frame image and a preset template face image, and calculating the similarity between each high-quality frame image and the template face image to obtain multiple high-dimensional similarity results; using a low-dimensional autoencoder to perform low-dimensional encoding on each high-quality frame image and a preset template face image, and calculating the similarity between each high-quality frame image and the template face image to obtain multiple low-dimensional similarity results; if all high-dimensional similarity results and low-dimensional similarity results are higher than the preset similarity threshold, the user identity is determined to be consistent with the user account; The speech recognition module is used to recognize the user's speech signal by using an end-to-end speech recognition model based on speech denoising and multiple recognition modules to obtain a speech recognition result if the verification result shows that the user identity is consistent with the user account; the recognition of the user's speech signal by using an end-to-end speech recognition model based on speech denoising and multiple recognition modules to obtain a speech recognition result includes: connecting multiple speech recognition network modules using different speech recognition algorithms in parallel, and connecting a speech denoising network module to the front end of the multiple speech recognition network modules, and connecting a speech denoising network module to the back end of the multiple speech recognition network modules. A speech recognition result comparison module is used to form an end-to-end speech recognition model; the end-to-end speech recognition model is used to recognize the user speech signal; the end-to-end speech recognition model is used to recognize the user speech signal, which includes: inputting the user speech signal into a speech denoising network module for denoising to obtain a denoised speech signal; inputting the denoised speech signal into multiple speech recognition network modules for speech recognition to obtain multiple recognition results; inputting the multiple recognition results into a speech recognition result comparison module for comparison and analysis, and outputting a final speech recognition result if the multiple recognition results are consistent; The concentration detection module is used to use an adaptive consumption-type concentration detection model based on classification confidence analysis to perform concentration detection on each frame image in the video signal to obtain multiple concentration detection results; it includes the following steps: selecting several facial images with high concentration as positive samples, and selecting several images of people with low concentration as negative samples; selecting some positive samples and some negative samples to train the SVM model to obtain a concentration detection model based on the SVM model; using all positive samples and all negative samples to train the convolutional neural network to obtain a concentration detection model based on the convolutional neural network; for any frame image, multi-scale reconstruction is performed to obtain images of multiple scales; using the concentration detection model based on the SVM model to detect images of multiple scales to obtain multiple detection results; if multiple detection results are consistent, the final concentration detection result is output; otherwise, the image is detected using the concentration detection model based on the convolutional neural network to obtain the final concentration detection result; The product recommendation module is used to determine and record products that the user is highly interested in based on the voice recognition results and multiple concentration detection results, and extract related products from the e-commerce platform system to form a product recommendation set and recommend it to the corresponding user; if the result of the user's high interest in the product being viewed at the time is obtained in any of the above voice recognition and concentration detection steps, the product being viewed by the user is directly recorded, and products of the same category and other highly related categories are recommended to the user for viewing.

3. An electronic device, characterized in that: include: a memory for storing one or more programs; processor; When the one or more programs are executed by the processor, the method according to claim 1 is implemented.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to claim 1 is implemented.

Images