CN-122025131-A - Pet health monitoring method, system, equipment and computer storage medium

Abstract

The application discloses a pet health monitoring method, a system, equipment and a computer storage medium, which relate to the technical field of data processing and are used for acquiring multi-mode original data acquired by a pet, preprocessing the multi-mode original data to obtain preprocessed data, extracting characteristics of the preprocessed data to obtain multi-mode characteristic data, and predicting health trend of the multi-mode characteristic data through a transducer model to obtain a health prediction state, wherein the multi-mode original data comprises physiological data, emotion data and environment data of the pet. The physiological data, the emotion data and the environmental data are all related to the health of the pet and reflect the health of the pet, so that the multi-mode original data can comprehensively reflect the health condition of the pet at multiple angles, and the transducer model can capture the health information of the pet from different representing subspaces, thereby realizing accurate health monitoring of the pet.

Inventors

• CHEN FANGXIONG
• MA XIAO

Assignees

• 深圳市广和通无线股份有限公司

Dates

Publication Date

20260512

Application Date

20260129

Claims (10)

1. 1. A method of pet health monitoring comprising: acquiring multi-mode original data obtained after the acquisition of the pets; Preprocessing the multi-mode original data to obtain preprocessed data; extracting features of the preprocessed data to obtain multi-mode feature data; carrying out health trend prediction on the multi-mode characteristic data through a transducer model to obtain a health prediction state; The multi-mode original data comprise physiological data, emotion data and environment data of the pet.
2. 2. The method for monitoring health of pets according to claim 1, wherein the predicting health trend of the multi-modal feature data by a transducer model to obtain a health prediction state further comprises: a service device that determines to serve the pet; constructing an action strategy according to the function of the service equipment; Constructing a state space based on the multimodal feature data and the health prediction state; Constructing a reward function positively related to the health of the pet according to the physiological data and the emotion data of the pet; Processing the action strategy, the state space and the rewarding function through reinforcement learning to generate a control strategy of the service equipment; and controlling the service equipment according to the control strategy.
3. 3. The pet health monitoring method of claim 2, wherein the service device comprises a pet feeding device, the processing the action strategy, the state space, and the reward function by reinforcement learning to generate a control strategy for the service device, comprising: processing the action strategy, the state space and the reward function through a depth deterministic strategy gradient algorithm to generate a pet feeding strategy; And generating a control strategy of the pet feeding device according to the pet feeding strategy.
4. 4. The method for monitoring health of pets according to claim 1, wherein the predicting health trend of the multi-modal feature data by a transducer model to obtain a health prediction state comprises: constructing a target time sequence feature according to the multi-mode feature data in the set time; Inputting the target time sequence characteristics into a pre-trained transducer model; Receiving a health prediction state output by a transducer model; The transducer model comprises a position coding adding layer, an encoder block and a classification head which are connected in sequence, wherein the position coding adding layer is used for adding position codes for time sequence features, the encoder block comprises a multi-head self-attention layer, a feedforward full-link layer and residual connection and layer normalization, and the classification head comprises a linear layer and a softmax layer.
5. 5. The method of claim 4, further comprising, prior to the predicting the health trend of the multimodal characterization data by the transducer model: acquiring training time sequence characteristics and health known states corresponding to the training time sequence characteristics; training the transducer model by applying the training time sequence characteristics; Generating a loss value of the transducer model through a loss function based on the health training state and the health known state output by the transducer model, wherein the loss function is constructed based on cross entropy loss and label smoothing; and adjusting the transducer model according to the loss value to obtain a trained transducer model.
6. 6. The method for monitoring health of pets according to claim 1, wherein the preprocessing of the multi-modal raw data to obtain preprocessed data comprises: acquiring a historical true value and a historical measured value which are obtained after the multi-mode sensor collects the pets; Performing least square fitting on the historical true value and the historical measured value to obtain a fitting result; determining acquisition error values of the multi-mode sensor according to the fitting result; and carrying out error correction on the multi-mode original data by using the acquired error value to obtain preprocessed data.
7. 7. The pet health monitoring method according to claim 1, wherein the feature extraction of the preprocessed data to obtain multi-modal feature data comprises: performing feature conversion on the weight of the pet in the pretreatment data to obtain weight feature data; processing the behavior action quantized value and the heart rate of the pet in the preprocessing data to generate activity level characteristic data; processing the emotion scores of the pets in the preprocessing data to generate emotion characteristic data; performing feature conversion on environment data of the pets in the pretreatment data to obtain environment feature data; and combining the weight characteristic data, the activity level characteristic data, the emotion characteristic data and the environment characteristic data into multi-modal characteristic data.
8. 8. A pet health monitoring system, comprising: the data acquisition module is used for acquiring multi-mode original data obtained after the acquisition of the pets; the preprocessing module is used for preprocessing the multi-mode original data to obtain preprocessed data; The feature extraction module is used for carrying out feature extraction on the preprocessed data to obtain multi-mode feature data; The health prediction module is used for predicting health trend of the multi-mode characteristic data through a transducer model to obtain a health prediction state; The multi-mode original data comprise physiological data, emotion data and environment data of the pet.
9. 9. An electronic device, comprising: A memory for storing a computer program; A processor for implementing the steps of the pet health monitoring method according to any one of claims 1 to 7 when executing the computer program.
10. 10. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein a computer program which, when executed by a processor, implements the steps of the pet health monitoring method of any of claims 1 to 7.

Description

Pet health monitoring method, system, equipment and computer storage medium Technical Field The present application relates to the field of data processing technology, and more particularly, to a method, a system, a device, and a computer storage medium for monitoring health of pets. Background In the maintenance process of pets such as cats and dogs, the health of the pets needs to be monitored independently in order to ensure the health of the pets and save the labor force of users. However, in the pet health monitoring process, the health condition of the pet cannot be accurately estimated, which is often performed according to the weight of the pet. In summary, how to accurately monitor the health of the pet is a problem to be solved by those skilled in the art. Disclosure of Invention The application aims to provide a pet health monitoring method which can solve the technical problem of accurately monitoring the health of pets to a certain extent. The application also provides a pet health monitoring system, electronic equipment and a computer readable storage medium. In order to achieve the above object, the present application provides the following technical solutions: a method of pet health monitoring comprising: acquiring multi-mode original data obtained after the acquisition of the pets; Preprocessing the multi-mode original data to obtain preprocessed data; extracting features of the preprocessed data to obtain multi-mode feature data; carrying out health trend prediction on the multi-mode characteristic data through a transducer model to obtain a health prediction state; The multi-mode original data comprise physiological data, emotion data and environment data of the pet. Preferably, the method further includes, after the health trend prediction is performed on the multimodal feature data by using a transducer model to obtain a health prediction state, the steps of: a service device that determines to serve the pet; constructing an action strategy according to the function of the service equipment; Constructing a state space based on the multimodal feature data and the health prediction state; Constructing a reward function positively related to the health of the pet according to the physiological data and the emotion data of the pet; Processing the action strategy, the state space and the rewarding function through reinforcement learning to generate a control strategy of the service equipment; and controlling the service equipment according to the control strategy. Preferably, the service device includes a pet feeding device, the processing the action policy, the state space and the reward function through reinforcement learning, generating a control policy of the service device includes: processing the action strategy, the state space and the reward function through a depth deterministic strategy gradient algorithm to generate a pet feeding strategy; And generating a control strategy of the pet feeding device according to the pet feeding strategy. Preferably, the performing health trend prediction on the multimodal feature data through a transducer model to obtain a health prediction state includes: constructing a target time sequence feature according to the multi-mode feature data in the set time; Inputting the target time sequence characteristics into a pre-trained transducer model; Receiving a health prediction state output by a transducer model; The transducer model comprises a position coding adding layer, an encoder block and a classification head which are connected in sequence, wherein the position coding adding layer is used for adding position codes for time sequence features, the encoder block comprises a multi-head self-attention layer, a feedforward full-link layer and residual connection and layer normalization, and the classification head comprises a linear layer and a softmax layer. Preferably, before the health trend prediction is performed on the multimodal feature data by using the transducer model, the method further includes: acquiring training time sequence characteristics and health known states corresponding to the training time sequence characteristics; training the transducer model by applying the training time sequence characteristics; Generating a loss value of the transducer model through a loss function based on the health training state and the health known state output by the transducer model, wherein the loss function is constructed based on cross entropy loss and label smoothing; and adjusting the transducer model according to the loss value to obtain a trained transducer model. Preferably, the preprocessing the multi-mode raw data to obtain preprocessed data includes: acquiring a historical true value and a historical measured value which are obtained after the multi-mode sensor collects the pets; Performing least square fitting on the historical true value and the historical measured value to obtain a fitting result; determining acquisition error values