AI digital human-computer interaction method, system, and medium applied to vending machines

Abstract

The present invention provides an AI digital human-computer interaction method, system, and medium applied to a vending machine, which belongs to the field of electronic digital data processing technology. The method includes: obtaining user facial information of a target user within a preset range of the vending machine; performing user analysis based on the user facial information and obtaining a user analysis description; obtaining voice data of the target user within a preset range; and performing AI digital human-computer interaction based on the voice data and the user analysis description. The AI ​​digital human-computer interaction method, system, and medium applied to a vending machine of the present invention obtain user facial information within a preset range of the vending machine, obtain user analysis description based on the user facial information, obtain voice data of the target user within a preset range, and perform AI digital human-computer interaction in combination with the voice data and the user analysis description, thereby improving the suitability of intelligent interaction and making the vending machine sales more intelligent.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic digital data processing, and in particular to an AI digital human-computer interaction method, system and medium applied to a vending machine. Background Art

[0002] AI digital human is a virtual being that uses artificial intelligence technology to simulate human appearance and behavior. It provides an interactive experience similar to that of a real person by simulating human facial expressions, body movements, voice patterns and emotional responses. At present, vending machines have enabled users to shop by themselves, but users still need to understand how the system operates before they can use it correctly. It is not smart enough and the user experience is poor.

[0003] The Chinese invention with application number: CN201810963942.8 discloses a vending machine transaction system with voice interaction, including a voice interaction module, a microphone, a speaker module, a first display screen capable of human-computer interaction, an analog signal and digital signal conversion module, a controller and an external power supply; the voice interaction module: includes a voice input unit and a voice processing unit, wherein the voice input unit is connected to the microphone and the voice processing unit, and the voice processing unit is connected to the controller; connected to the controller controlling the vending machine through the analog signal and digital signal conversion module; used for human-computer interaction and playing product information; the analog signal and digital signal conversion module: connected to the controller and the first display screen at the same time. On the one hand, the above invention can realize voice interaction with users to better meet the needs of users, and on the other hand, it can timely and effectively remind users when new products are put on the shelves and push out corresponding products according to the actual needs of users.

[0004] However, when users shop through vending machines, there are situations where the shopping demands they make are inappropriate (for example, excessive demands), resulting in waste. Direct reminders based on user voice interaction without considering the user's own situation are not smart enough, and the interaction is not targeted and has poor suitability.

[0005] In view of this, there is an urgent need for AI digital human-computer interaction methods, systems, and media applied to vending machines to at least solve the above-mentioned deficiencies. Summary of the invention

[0006] One of the purposes of the present invention is to provide an AI digital human-computer interaction method, system, and medium for use in vending machines, to obtain user facial information within a preset range of the vending machine, to obtain a user analysis description based on the user facial information, to obtain voice data of a target user within a preset range, and to perform AI digital human-computer interaction in combination with the voice data and the user analysis description, thereby improving the suitability of intelligent interaction and making the vending machine's sales more intelligent.

[0007] The AI ​​digital human-computer interaction method applied to a vending machine provided in an embodiment of the present invention includes:

[0008] Obtaining user facial information of a target user within a preset range of the vending machine;

[0009] Perform user analysis based on user facial information and obtain user analysis description;

[0010] Acquire the target user's voice data within a preset range;

[0011] Perform AI digital human-computer interaction based on voice data and user analysis description.

[0012] Preferably, obtaining the user facial information of a target user within a preset range of the vending machine includes:

[0013] Based on AI visual analysis technology, capture and analyze the user face of the target user;

[0014] According to the user's face and the historical user face database preset by the vending machine, the target user is judged as a new or old user, and the judgment result of the new or old user is obtained, and the judgment result includes: the user's face marked with a new user label and an old user label;

[0015] The determination result is used as the user's facial information.

[0016] Preferably, performing user analysis based on user facial information to obtain a user analysis description includes:

[0017] Parse the user's facial information and obtain the user's tag type;

[0018] If the user tag type is a new user, the user analysis basis is obtained according to the corresponding user face;

[0019] Obtaining a first analysis model preset according to the type;

[0020] Obtaining a user analysis description according to the user analysis basis and the first analysis model;

[0021] If the user tag type is an old user, then the corresponding user face is matched with the historically captured user face library to retrieve the user's historical behavior;

[0022] Based on the preset shopping behavior feature extraction template, shopping behavior features are extracted according to the user's historical behavior;

[0023] Obtain a second analysis model preset for the user's historical behavior;

[0024] According to the shopping behavior characteristics and the second analysis model, a user analysis description is obtained.

[0025] Preferably, AI digital human-computer interaction is performed based on the voice data and user analysis description, including:

[0026] Analyze voice data to determine whether there are active trigger keywords;

[0027] If there is an active trigger keyword, AI digital human-computer interaction is performed based on the target voice data and user analysis description after the active trigger keyword;

[0028] If there are no active trigger keywords, AI digital human-computer interaction will be performed based on the user analysis description.

[0029] Preferably, according to the target voice data and user analysis description after the active triggering of the keyword, AI digital human-computer interaction is performed, including:

[0030] Analyze the target voice data and extract voice keywords;

[0031] Get the product keywords of the products sold in the vending machine;

[0032] Traversing the voice keywords in sequence, matching the voice keyword being traversed with each product keyword in sequence, and determining whether they match consistently. If they match consistently, obtaining the first keyword position of the voice keyword in the voice keywords that corresponds to the consistently matched voice keyword;

[0033] Determine the voice keywords of preset keyword position units before and after the first keyword position as the second keyword;

[0034] Determine the quantity information of the product according to the second keyword and the preset quantity keyword library;

[0035] According to the second keyword and the preset commodity confirmation attitude keyword library, try to obtain commodity confirmation attitude information;

[0036] Determine the purchase product information according to the voice keyword corresponding to the first keyword position, the product quantity information and the product confirmation attitude information;

[0037] Based on the purchased product information and the user analysis description, conduct a purchase rationality analysis on the purchased product information and obtain the purchase rationality analysis result;

[0038] Based on a preset purchase rationality analysis result feature extraction template, extract purchase rationality analysis result features according to the purchase rationality analysis result;

[0039] Determine the human-computer interaction strategy based on the characteristics of the purchase rationality analysis results;

[0040] Conduct AI digital human-computer interaction based on human-computer interaction strategies.

[0041] Preferably, according to the user analysis description, AI digital human-computer interaction is performed, including:

[0042] Extract the passive trigger basis according to the user analysis description and the preset passive trigger basis extraction template;

[0043] If the extraction is successful, the human-computer interaction strategy is determined based on the passive trigger basis and the preset human-computer interaction library, and AI digital human-computer interaction is performed according to the human-computer interaction strategy.

[0044] The AI ​​digital human-computer interaction method for a vending machine provided in an embodiment of the present invention further includes:

[0045] The inventory information of the goods being sold is obtained, and when there is insufficient inventory, alternative goods for sale are recommended; wherein insufficient inventory means that the quantity of goods corresponding to the quantity information of the goods is greater than the quantity in stock.

[0046] Preferably, when there is insufficient inventory, alternative products for sale are recommended, including:

[0047] Acquire first commodity attribute information of the sold commodity with insufficient inventory, quantize the first commodity attribute information based on the first commodity attribute feature vector, and acquire the first commodity description vector; wherein the first commodity attribute feature is extracted from the first commodity attribute information based on a preset feature extraction template;

[0048] Acquire the second commodity attribute information of the remaining commodities in the vending machine except for the commodities that are out of stock, and quantize the second commodity attribute information based on the second commodity attribute feature vector to acquire the second commodity description vector;

[0049] The first product description vector and the second product description vector are matched, and the remaining sales products described by the second product description vector with the largest vector matching degree are selected as replacement products.

[0050] The AI ​​digital human-computer interaction system for vending machines provided in an embodiment of the present invention includes:

[0051] A user face information acquisition subsystem, used to acquire the user face information of a target user within a preset range of the vending machine;

[0052] A user analysis description acquisition subsystem, used to perform user analysis based on user facial information and acquire user analysis description;

[0053] The voice data collection subsystem is used to obtain the voice data of the target user within a preset range;

[0054] The human-computer interaction subsystem is used to perform AI digital human-computer interaction based on voice data and user analysis description.

[0055] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the computer program implements any step of the above-mentioned AI digital human-computer interaction method.

[0056] The beneficial effects of the present invention are:

[0057] The present invention obtains user facial information within a preset range of a vending machine, obtains a user analysis description based on the user facial information, obtains voice data of a target user within a preset range, and combines the voice data and the user analysis description to perform AI digital human-computer interaction, thereby improving the suitability of intelligent interaction and making the vending machine's sales more intelligent.

[0058] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures specifically pointed out in the present application documents.

[0059] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0061] Figure 1 Schematic diagram of an AI digital human-computer interaction method applied to a vending machine in an embodiment of the present invention;

[0062] Figure 2 Schematic diagram of an AI digital human-computer interaction system applied to a vending machine in an embodiment of the present invention. DETAILED DESCRIPTION

[0063] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0064] The embodiment of the present invention provides an AI digital human-computer interaction method applied to a vending machine, such as Figure 1 As shown, including:

[0065] Step 1: Obtain the user facial information of the target user within the preset range of the vending machine; wherein the preset range is preset manually, for example: a ray is drawn with the location of the vending machine as the starting point and the display direction of the vending machine (perpendicular to the display window of the vending machine) and a semicircular area with a radius of 1.5 meters bisected by the ray is set as the preset range; the user facial information of the target user is: the facial information (face ID) recognized by the face recognition device on the vending machine within the preset range, and the target user is the recognized person corresponding to the facial information;

[0066] Step 2: Perform user analysis based on user facial information to obtain user analysis description; user analysis is: analyzing the user's shopping preference analysis behavior, for example: obtaining the user's gender and age group information based on face recognition, and then determining the shopping preference based on the shopping habits of the corresponding gender and age group information; for example: obtaining the user's historical shopping habits based on face recognition, and determining the shopping preference based on the historical shopping habits; user analysis description is: constructing a user shopping preference description vector based on the collected user data (such as facial information, purchase history, behavioral habits, etc.);

[0067] Step 3: Acquire the voice data of the target user within a preset range; wherein the voice data is: the sound data of the target user recorded by the microphone mounted on the vending machine;

[0068] Step 4: Perform AI digital human-computer interaction based on voice data and user analysis description. When AI digital human-computer interaction is performed based on voice data and user analysis description, the AI ​​digital human understands the semantics of the voice data and conducts a real-time dialogue with the target user. During the dialogue, if the shopping demand information in the target user's voice data is clear, a confirmation dialogue of the shopping demand information is directly output. Otherwise, based on the user analysis description and voice data, the shopping demand information of the target user is first predicted and then a confirmation dialogue of the predicted shopping demand information is output or a shopping recommendation is made to the target user. The confirmation dialogue or shopping recommendation dialogue is replied by the AI ​​digital human.

[0069] The working principle and beneficial effects of the above technical solution are:

[0070] The present invention obtains user facial information within a preset range of a vending machine, obtains a user analysis description based on the user facial information, obtains voice data of a target user within a preset range, and combines the voice data and the user analysis description to perform AI digital human-computer interaction, thereby improving the suitability of intelligent interaction and making the vending machine's sales more intelligent.

[0071] In one embodiment, obtaining user facial information of a target user within a preset range of a vending machine includes:

[0072] Based on AI visual analysis technology, capture and analyze the user face of the target user;

[0073] According to the user face and the historical user face library preset by the vending machine, the target user is judged as a new or old user, and the judgment result of the new or old user is obtained, and the judgment result includes: the user face marked with a new user label and an old user label; wherein the historical user face library preset by the vending machine includes: the face ID of the historical user who successfully purchased the commodity and was captured by the vending machine in the past; the new or old user is judged as follows: if there is a user face that matches any historical user face in the historical user face library, then the old user label is marked for the corresponding user face, otherwise, the new user label is marked for the corresponding user face;

[0074] The determination result is used as the user's facial information.

[0075] The working principle and beneficial effects of the above technical solution are:

[0076] The present invention introduces AI visual analysis technology to obtain the user face of the target user, introduces a historically captured user face library preset in the vending machine, matches the user face with the historically captured user faces, determines the user faces with matching old user tags and the user faces with non-matching new user tags, and uses the determination result of the new and old user faces as the user face information, so that the acquisition of user face information is more reasonable.

[0077] In one embodiment, performing user analysis based on user facial information to obtain a user analysis description includes:

[0078] Parse the user's facial information and obtain the user's tag type;

[0079] If the user tag type is a new user, the user analysis basis is obtained according to the corresponding user face; the user analysis basis is: the basis for analyzing the user shopping behavior of the new user, such as: the age group is children, the gender is female;

[0080] Obtain a first analysis model preset based on the type; wherein the first analysis model is: an AI model for analyzing the shopping habits of users containing the basis of the type, such as: an AI model for analyzing the shopping habits of users of different age groups. When obtaining the first analysis model, obtain an artificial analysis record of analyzing the shopping habits of users based on the basis of the type through big data (such as: an artificial analysis record of analyzing the shopping habits of users based on age), use the basis of the type in the artificial analysis record as input, and use the description vector of the user's shopping habits obtained from the artificial analysis record as output to train a neural network model to obtain the first analysis model;

[0081] Obtaining a user analysis description according to the user analysis basis and the first analysis model; wherein the user analysis basis is input into the first analysis model, and the corresponding output of the first analysis model is the user analysis description;

[0082] If the user tag type is an old user, then the corresponding user face is matched with a matching historically captured user face in the historically captured user face database, and the user's historical behavior is retrieved; wherein the user's historical behavior specifically includes: the old user's historical shopping behavior;

[0083] Based on the preset shopping behavior feature extraction template, shopping behavior features are extracted according to the user's historical behavior; wherein the preset shopping behavior feature extraction template is a template for extracting shopping behavior features according to the user's historical behavior, for example, if the shopping behavior features are shopping time feature values ​​(for example, 18:00 every day) and shopping category feature values ​​(for example, bottled water), then the shopping behavior feature extraction template is a template for extracting shopping time information and shopping category information according to the user's historical behavior;

[0084] Acquire a second analysis model preset for the user's historical behavior; wherein the second analysis model preset for the user's historical behavior is: an AI model for analyzing the shopping habits of users with different historical shopping behaviors, and when acquiring the second analysis model, obtain an analysis record of manually analyzing the user's shopping habits based on the historical shopping behaviors, use the behavioral features of the historical shopping behaviors as input, and use the description vectors describing the shopping habits obtained by the analysis as output to train a CNN model, and obtain a second analysis model trained to convergence;

[0085] According to the shopping behavior characteristics and the second analysis model, a user analysis description is obtained.

[0086] The working principle and beneficial effects of the above technical solution are:

[0087] For new and old users, this embodiment adopts different methods to obtain user analysis descriptions. When the user is a new user, since there is no historical purchase record of the user, it is impossible to predict shopping habits based on historical shopping habits. Therefore, the user analysis basis is obtained based on the recognized user face, and the first analysis model preset for the basis type of the user analysis basis is introduced (an AI model that analyzes the shopping habits of users with the basis type), and the user analysis description is obtained based on the user analysis basis and the first analysis model. When the user is an old user, it is easy to retrieve the user's historical behavior based on the old user's face; the shopping behavior feature extraction template is introduced to extract the shopping behavior features in the user's historical behavior; the second analysis model preset for the user's historical behavior is obtained (an AI model that analyzes the shopping habits of users with different historical shopping behaviors), and the shopping behavior features are input into the second analysis model to obtain the user analysis description. The user analysis description is obtained based on the different adaptability analysis of new and old users, which improves the rationality and accuracy of obtaining the user analysis description.

[0088] In one embodiment, AI digital human-computer interaction is performed based on voice data and user analysis description, including:

[0089] Parse the voice data to determine whether there are active trigger keywords; parsing the voice data means converting the voice data into semantic text data based on voice recognition technology; active trigger keywords are preset manually, such as "I want to buy...", "I want to buy...";

[0090] If there is an active trigger keyword, AI digital human-computer interaction is performed based on the target voice data after the active trigger keyword and the user analysis description; wherein the target voice data after the active trigger keyword is the voice data after the position of the active trigger keyword is extracted from the voice data;

[0091] If there are no active trigger keywords, AI digital human-computer interaction will be performed based on the user analysis description.

[0092] The working principle and beneficial effects of the above technical solution are:

[0093] The present invention introduces active trigger keywords, parses voice data and attempts to identify active trigger keywords. When active trigger keywords exist, it indicates that the user actively initiates a purchase request. AI digital human-computer interaction is performed based on the target voice data after the active trigger keywords and the user analysis description. Otherwise, the interaction is directly performed based on the user analysis description, thereby improving the rationality of the interaction logic.

[0094] In one embodiment, according to the target voice data and user analysis description after the active triggering of the keyword, AI digital human-computer interaction is performed, including:

[0095] Parse the target voice data and extract voice keywords; wherein the method of extracting voice keywords is the same as the principle of extracting active trigger keywords;

[0096] Obtain product keywords of products sold in the vending machine; wherein the product keywords are: semantics describing the products sold in the vending machine;

[0097] The voice keywords are traversed in sequence, and the voice keywords being traversed are matched with each product keyword in sequence to determine whether they match each other. If they match each other, the first keyword position of the voice keyword that matches each other in the voice keywords is obtained; wherein the first keyword position is: the position of the voice keyword that matches the product keyword in the voice keyword sequence in which it is located;

[0098] Determine the voice keywords of preset keyword position units before and after the first keyword position as the second keyword; wherein the preset keyword position units are set by the staff, for example: 5;

[0099] Determine the quantity information of the commodity according to the second keyword and a preset quantity keyword library; wherein the preset quantity keyword library is a database composed of keywords describing quantity;

[0100] According to the second keyword and the preset commodity confirmation attitude keyword library, try to obtain commodity confirmation attitude information; wherein the preset commodity confirmation attitude keyword library is: a database composed of keywords describing the commodity confirmation status, and the commodity confirmation attitude keywords are, for example: "need", "previously...not needed";

[0101] Determine the purchase product information according to the voice keyword corresponding to the first keyword position, the product quantity information and the product confirmation attitude information;

[0102] According to the purchased commodity information and the user analysis description, a purchase rationality analysis of the purchased commodity information is performed to obtain the purchase rationality analysis result; wherein, the purchase rationality analysis of the purchased commodity information refers to: the process of inputting the purchased commodity information and the user analysis description into the purchase rationality analysis model to perform a rationality analysis of the user's purchase demand; the purchase rationality analysis result includes: reasonable, unreasonable and unreasonable reasons; the purchase rationality analysis model is: parsing the manual records of the manual rationality analysis of the user's purchase of commodities, determining the purchase information and the user shopping behavior description input into the preset neural network model, outputting the manual rationality analysis result of the preset neural network model and performing model training, and when the above neural network model is trained to convergence, the purchase rationality analysis model is obtained;

[0103] Based on the preset purchase rationality analysis result feature extraction template, according to the purchase rationality analysis result, the purchase rationality analysis result feature is extracted; wherein the purchase rationality analysis result feature is: rationality, irrationality and irrationality reason; the preset purchase rationality analysis result feature extraction template is used as a template for extracting purchase rationality analysis result features according to the purchase rationality analysis result, for example: extracting irrationality and irrationality reason according to the purchase rationality analysis result. System execution rules;

[0104] Determine the human-computer interaction strategy according to the characteristics of the purchase rationality analysis result; wherein, the human-computer interaction strategy determined according to the characteristics of the purchase rationality analysis result is determined based on the matching of the historical strategy library, the historical strategy library includes one-to-one correspondence between pre-selected rationality analysis result characteristics and corresponding pre-selected human-computer interaction strategies, and the historical strategy library is manually preset based on experience and distribution of multiple purchase rationality analysis result characteristics;

[0105] Conduct AI digital human-computer interaction based on human-computer interaction strategies.

[0106] The working principle and beneficial effects of the above technical solution are:

[0107] The present invention extracts voice keywords from target voice data and simultaneously obtains commodity keywords of commodities sold in a vending machine; sequentially traverses the voice keywords, sequentially matches the voice keywords being traversed with each commodity keyword, determines the first keyword position of the matched voice keyword in the voice keywords, and obtains the voice keywords of preset keyword position units before and after the first keyword position as second keywords;

[0108] A keyword library consisting of keywords describing quantity is introduced, and keyword matching is performed based on the second keyword and the quantity keyword library to determine the quantity information of the product; a product confirmation attitude keyword library describing the purchase attitude of the product is introduced, and keyword matching is performed based on the second keyword and the product confirmation attitude keyword library to determine the product confirmation attitude information; the voice keyword corresponding to the first keyword position, the product quantity information and the product confirmation attitude information are combined to determine the purchase product information, and AI digital human-computer interaction is performed (for example: the purchase product information details are displayed on the display screen of the vending machine for the target user to confirm. After confirmation, the vending machine executes the shipment instruction, and the target user does not need to operate the vending system, which has a low learning cost and is more convenient;

[0109] Furthermore, when AI digital humans interact with each other based on the purchased goods information, a purchase rationality analysis of the purchased goods information is performed, and a preset purchase rationality analysis result feature extraction template is introduced to extract the purchase rationality analysis result features in the purchase rationality analysis results. Based on the purchase rationality analysis result features, a human-computer interaction strategy is determined and AI digital humans interact with each other. For example, if the purchase rationality analysis result features are unreasonable and the unreasonable reason is that the purchase exceeds the historical average purchase quantity by 2 times, then the human-computer interaction strategy is to remind the purchasing user that the purchase may be over-purchasing, and then interact based on the human-computer interaction strategy. The human-computer interaction strategy is determined in combination with the actual purchase situation and user habits to interact, thereby improving the suitability of AI interaction.

[0110] In one embodiment, according to the user analysis description, AI digital human-computer interaction is performed, including:

[0111] According to the user analysis description and the preset passive trigger basis extraction template, the passive trigger basis is extracted; wherein the preset passive trigger basis extraction template is used for extracting the passive trigger basis in accordance with the user analysis description, and the triggered basis is: a feature that can directly trigger the vending machine to execute the instruction without the target user actively inputting the instruction to the vending machine, for example: user A has purchased a can of coffee at 2 pm every day in the past, then it is detected that user A walks into the preset range at 2 pm, which is the triggered basis, and the corresponding triggered instruction is: generating a pending order for purchasing a can of coffee for user A to view;

[0112] If the extraction is successful, the human-computer interaction strategy is determined based on the passive trigger basis and the preset human-computer interaction library, and the AI ​​digital human-computer interaction is performed according to the human-computer interaction strategy. Among them, the preset human-computer interaction library includes one-to-one corresponding passive trigger basis to be selected and human-computer interaction strategy to be confirmed.

[0113] The working principle and beneficial effects of the above technical solution are:

[0114] The present invention introduces a passive trigger basis extraction template, extracts the passive trigger basis according to the user analysis description, introduces a human-computer interaction library, determines the human-computer interaction strategy for passive triggering of the target user, and performs AI digital human-computer interaction based on the human-computer interaction strategy. When the target user meets the characteristic conditions of passive triggering, adaptive recommendations will be made without the user's active use, thereby improving the success rate of commodity sales.

[0115] The embodiment of the present invention provides an AI digital human-computer interaction method applied to a vending machine, further comprising:

[0116] Acquire inventory information of the sold goods, and recommend substitute goods for the sold goods when there is insufficient inventory; insufficient inventory means that the quantity of goods corresponding to the quantity information of the goods is greater than the quantity in stock; the inventory information includes remaining quantity information of the sold goods, and when recommending substitute goods, calculate the goods with the greatest similarity to the goods attributes of the sold goods according to the pre-entered goods attribute information of each good as substitute goods, and display the substitute goods information to the user for viewing, so as to facilitate the user to make substitute choices;

[0117] When there is insufficient inventory, we recommend alternative products for sale, including:

[0118] Acquire the first commodity attribute information of the sold commodity with insufficient inventory, quantize the first commodity attribute information based on the first commodity attribute feature vector, and acquire the first commodity description vector; wherein, the first commodity attribute feature is extracted according to the first commodity attribute information based on a preset feature extraction template; the first commodity attribute information is: commodity attribute data of the sold commodity with insufficient inventory, including: commodity variety, brand, price and ingredients, etc., and the first commodity attribute information is acquired through the entry record when the sold commodity is put into storage; the first commodity attribute features include: commodity variety feature value (for example: potato chips), brand feature value (for example: certain brand), price feature value (for example: 12.99 yuan) and ingredient feature value (for example: potatoes, salt), etc.; the preset feature extraction template is: a template for extracting commodity attribute features of different feature types according to the commodity attribute information, such as: a rule for extracting commodity variety feature values ​​according to the record of the first commodity attribute information; the construction rule for constructing the commodity description vector according to the commodity attribute features is preset manually;

[0119] Acquire the second commodity attribute information of the remaining commodities in the vending machine except for the commodities that are out of stock, quantize the second commodity attribute information based on the second commodity attribute feature vector, and acquire the second commodity description vector; wherein the second commodity attribute information is: commodity attribute data of the remaining commodities; the second commodity description vector is constructed based on the second commodity attribute feature extracted by the feature extraction template, and the second commodity attribute information and the second commodity description vector are acquired in the same manner as the first commodity attribute information and the first commodity description vector;

[0120] The first product description vector and the second product description vector are matched, and the remaining products described by the second product description vector with the largest vector matching degree are selected as replacement products. The vector matching degree refers to the vector cosine similarity.

[0121] The working principle and beneficial effects of the above technical solution are:

[0122] The present invention introduces the inventory information of the sold goods. When the inventory is insufficient, the first commodity attribute information entered when the corresponding sold goods are entered into the warehouse is obtained, the first commodity attribute information is characterized to obtain the first commodity attribute feature, and a first commodity description vector is constructed based on the first commodity attribute feature; the second commodity attribute information of the sold goods other than the sold goods with insufficient inventory is obtained, and a second commodity description vector is constructed based on the same construction method as the first commodity description vector, the first commodity description vector and the second commodity description vector are matched, and the remaining sold goods described by the corresponding matching second commodity description vector are determined as substitute goods. When the inventory of the goods required by the user is insufficient, substitute goods are recommended in time, and substitute goods with similar commodity attributes are recommended in time, so as to enrich the choices of the target users and improve the success rate of commodity sales.

[0123] The embodiment of the present invention provides an AI digital human-computer interaction system applied to a vending machine, such as Figure 2 As shown, including:

[0124] User facial information acquisition subsystem 1, used to acquire user facial information of a target user within a preset range of the vending machine;

[0125] User analysis description acquisition subsystem 2, used to perform user analysis based on user facial information and acquire user analysis description;

[0126] The voice data collection subsystem 3 is used to obtain the voice data of the target user within a preset range;

[0127] The human-computer interaction subsystem 4 is used to perform AI digital human-computer interaction based on voice data and user analysis description.

[0128] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the computer program implements any step of the above-mentioned AI digital human-computer interaction method.

[0129] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. An AI digital human-computer interaction method applied to vending machines, characterized in that: include: Obtaining user facial information of a target user within a preset range of the vending machine; wherein the preset range is a semicircular area with a radius of 1.5 meters bisected by a ray constructed from the location of the vending machine and the display direction of the vending machine; Perform user analysis based on user facial information to obtain a user analysis description; wherein the user analysis description is: a user shopping preference description vector constructed based on the collected user data; The performing of user analysis based on user facial information and obtaining a user analysis description includes: Parse the user's facial information and obtain the user's tag type; If the user tag type is a new user, the user analysis basis is obtained according to the corresponding user face; Obtain a first analysis model preset based on the type; wherein the first analysis model is: an AI model for analyzing the shopping habits of a user including the basis of the type, when obtaining the first analysis model, obtain a manual analysis record of analyzing the shopping habits of the user based on the basis of the type through big data, use the basis of the type in the manual analysis record as input, and use the description vector of the user's shopping habits obtained from the manual analysis record as output to train a neural network model to obtain the first analysis model; Obtaining a user analysis description according to the user analysis basis and the first analysis model; Acquire the target user's voice data within a preset range; Conduct AI digital human-computer interaction based on voice data and user analysis description; The AI ​​digital human-computer interaction is performed based on the voice data and user analysis description, including: Analyze voice data to determine whether there are active trigger keywords; If there is an active trigger keyword, AI digital human-computer interaction is performed based on the target voice data and user analysis description after the active trigger keyword; If there are no active trigger keywords, AI digital human-computer interaction is performed based on the user analysis description; The method of performing AI digital human-computer interaction based on the target voice data and user analysis description after the active triggering of the keyword includes: Analyze the target voice data and extract voice keywords; Get the product keywords of the products sold in the vending machine; Traversing the voice keywords in sequence, matching the voice keyword being traversed with each product keyword in sequence, and determining whether they match consistently. If they match consistently, obtaining the first keyword position of the voice keyword in the voice keywords that corresponds to the consistently matched voice keyword; Determine the voice keywords of preset keyword position units before and after the first keyword position as the second keyword; Determine the quantity information of the product according to the second keyword and the preset quantity keyword library; According to the second keyword and the preset commodity confirmation attitude keyword library, try to obtain commodity confirmation attitude information; Determine the purchase product information according to the voice keyword corresponding to the first keyword position, the product quantity information and the product confirmation attitude information; According to the purchased commodity information and the user analysis description, a purchase rationality analysis of the purchased commodity information is performed to obtain the purchase rationality analysis result; wherein, the purchase rationality analysis of the purchased commodity information refers to: the process of inputting the purchased commodity information and the user analysis description into the purchase rationality analysis model to perform a rationality analysis of the user's purchase demand; the purchase rationality analysis result includes: reasonable, unreasonable and unreasonable reasons; the purchase rationality analysis model is: parsing the manual records of the manual rationality analysis of the user's purchase of commodities, determining the purchase information and the user shopping behavior description input into the preset neural network model, outputting the manual rationality analysis result of the preset neural network model and performing model training, and when the above neural network model is trained to convergence, the purchase rationality analysis model is obtained; Based on the preset purchase rationality analysis result feature extraction template, according to the purchase rationality analysis result, the purchase rationality analysis result feature is extracted; wherein the purchase rationality analysis result feature is: rationality, irrationality and irrationality reason; the preset purchase rationality analysis result feature extraction template is used as a template for extracting purchase rationality analysis result features according to the purchase rationality analysis result, and the system execution rules for extracting irrationality and irrationality reason according to the purchase rationality analysis result; Determine the human-computer interaction strategy according to the characteristics of the purchase rationality analysis result; wherein, the human-computer interaction strategy determined according to the characteristics of the purchase rationality analysis result is determined based on the matching of the historical strategy library, the historical strategy library includes one-to-one correspondence between pre-selected rationality analysis result characteristics and corresponding pre-selected human-computer interaction strategies, and the historical strategy library is manually preset based on experience and distribution of multiple purchase rationality analysis result characteristics; Conduct AI digital human-computer interaction based on human-computer interaction strategies.

2. The AI ​​digital human-computer interaction method for vending machines according to claim 1, characterized in that: Obtaining user facial information of a target user within a preset range of the vending machine, including: Based on AI visual analysis technology, capture and analyze the user face of the target user; According to the user's face and the historical user face database preset by the vending machine, the target user is judged as a new or old user, and the judgment result of the new or old user is obtained, and the judgment result includes: the user's face marked with a new user label and an old user label; The determination result is used as the user's facial information.

3. The AI ​​digital human-computer interaction method for vending machines according to claim 1, characterized in that: Perform user analysis based on user facial information and obtain user analysis description, which also includes: If the user tag type is an old user, then the corresponding user face is matched with the historically captured user face library to retrieve the user's historical behavior; Based on the preset shopping behavior feature extraction template, shopping behavior features are extracted according to the user's historical behavior; Obtain a second analysis model preset for the user's historical behavior; According to the shopping behavior characteristics and the second analysis model, a user analysis description is obtained.

4. The AI ​​digital human-computer interaction method for a vending machine as claimed in claim 1, characterized in that: According to the user analysis description, AI digital human-computer interaction is carried out, including: Extract the passive trigger basis according to the user analysis description and the preset passive trigger basis extraction template; If the extraction is successful, the human-computer interaction strategy is determined based on the passive trigger basis and the preset human-computer interaction library, and AI digital human-computer interaction is performed according to the human-computer interaction strategy.

5. The AI ​​digital human-computer interaction method for vending machines according to claim 1, characterized in that: Also includes: Get the inventory information of the goods being sold, and recommend alternative goods when there is insufficient inventory; insufficient inventory means: The product quantity corresponding to the product quantity information is greater than the inventory quantity.

6. The AI ​​digital human-computer interaction method for a vending machine as claimed in claim 5, characterized in that: When there is insufficient inventory, we recommend alternative products for sale, including: Acquire first commodity attribute information of the sold commodity with insufficient inventory, quantize the first commodity attribute information based on the first commodity attribute feature vector, and acquire the first commodity description vector; wherein the first commodity attribute feature is extracted from the first commodity attribute information based on a preset feature extraction template; Acquire the second commodity attribute information of the remaining commodities in the vending machine except for the commodities that are out of stock, and quantize the second commodity attribute information based on the second commodity attribute feature vector to acquire the second commodity description vector; The first product description vector and the second product description vector are matched, and the remaining sales products described by the second product description vector with the largest vector matching degree are selected as replacement products.

7. AI digital human-computer interaction system applied to vending machines, characterized in that: include: A user facial information acquisition subsystem is used to acquire user facial information of a target user within a preset range of the vending machine; wherein the preset range is a semicircular area with a radius of 1.5 meters bisected by a ray constructed from the location of the vending machine and the display direction of the vending machine; The user analysis description acquisition subsystem is used to perform user analysis based on user facial information and obtain user analysis description; wherein the user analysis description is: a user shopping preference description vector constructed based on the collected user data; The user analysis description acquisition subsystem performs user analysis based on user facial information and obtains user analysis descriptions, including: Parse the user's facial information and obtain the user's tag type; If the user tag type is a new user, the user analysis basis is obtained according to the corresponding user face; Obtain a first analysis model preset based on the type; wherein the first analysis model is: an AI model for analyzing the shopping habits of a user including the basis of the type, when obtaining the first analysis model, obtain a manual analysis record of analyzing the shopping habits of the user based on the basis of the type through big data, use the basis of the type in the manual analysis record as input, and use the description vector of the user's shopping habits obtained from the manual analysis record as output to train a neural network model to obtain the first analysis model; According to the user analysis basis and the first analysis model, a user analysis description is obtained; a voice data collection subsystem is used to obtain the voice data of the target user within a preset range; The human-computer interaction subsystem is used to perform AI digital human-computer interaction based on voice data and user analysis description; The human-computer interaction subsystem performs AI digital human-computer interaction based on voice data and user analysis description, including: Analyze voice data to determine whether there are active trigger keywords; If there is an active trigger keyword, AI digital human-computer interaction is performed based on the target voice data and user analysis description after the active trigger keyword; If there are no active trigger keywords, AI digital human-computer interaction is performed based on the user analysis description; The method of performing AI digital human-computer interaction based on the target voice data and user analysis description after the active triggering of the keyword includes: Analyze the target voice data and extract voice keywords; Get the product keywords of the products sold in the vending machine; Traversing the voice keywords in sequence, matching the voice keyword being traversed with each product keyword in sequence, and determining whether they match consistently. If they match consistently, obtaining the first keyword position of the voice keyword in the voice keywords that corresponds to the consistently matched voice keyword; Determine the voice keywords of preset keyword position units before and after the first keyword position as the second keyword; Determine the quantity information of the product according to the second keyword and the preset quantity keyword library; According to the second keyword and the preset commodity confirmation attitude keyword library, try to obtain commodity confirmation attitude information; Determine the purchase product information according to the voice keyword corresponding to the first keyword position, the product quantity information and the product confirmation attitude information; According to the purchased commodity information and the user analysis description, a purchase rationality analysis of the purchased commodity information is performed to obtain the purchase rationality analysis result; wherein, the purchase rationality analysis of the purchased commodity information refers to: the process of inputting the purchased commodity information and the user analysis description into the purchase rationality analysis model to perform a rationality analysis of the user's purchase demand; the purchase rationality analysis result includes: reasonable, unreasonable and unreasonable reasons; the purchase rationality analysis model is: parsing the manual records of the manual rationality analysis of the user's purchase of commodities, determining the purchase information and the user shopping behavior description input into the preset neural network model, outputting the manual rationality analysis result of the preset neural network model and performing model training, and when the above neural network model is trained to convergence, the purchase rationality analysis model is obtained; Based on the preset purchase rationality analysis result feature extraction template, according to the purchase rationality analysis result, the purchase rationality analysis result feature is extracted; wherein the purchase rationality analysis result feature is: rationality, irrationality and irrationality reason; the preset purchase rationality analysis result feature extraction template is used as a template for extracting purchase rationality analysis result features according to the purchase rationality analysis result, and the system execution rules for extracting irrationality and irrationality reason according to the purchase rationality analysis result; Determine the human-computer interaction strategy according to the characteristics of the purchase rationality analysis result; wherein, the human-computer interaction strategy determined according to the characteristics of the purchase rationality analysis result is determined based on the matching of the historical strategy library, the historical strategy library includes one-to-one correspondence between pre-selected rationality analysis result characteristics and corresponding pre-selected human-computer interaction strategies, and the historical strategy library is manually preset based on experience and distribution of multiple purchase rationality analysis result characteristics; Conduct AI digital human-computer interaction based on human-computer interaction strategies.

8. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of the AI ​​digital human-computer interaction method according to any one of claims 1 to 6 are implemented.

Images