| CPC G06F 11/3438 (2013.01) [G06F 21/602 (2013.01); G06N 7/01 (2023.01); G06N 20/00 (2019.01)] | 2 Claims |
|
1. An activity recognition system, comprising a communication framework jointly formed by at least one data collecting terminal, each of the at least one edge computing devices, and a cloud computing platform, the activity recognition system uses the communication framework to conduct personnel activity recognition and model updating, wherein at least one edge computing device at least comprises:
a data pre-processing module, for pre-processing sensing signals collected by the at least one data collecting terminal so as to obtain first data,
wherein the data pre-processing module uses pre-processing techniques including at least one or more of calibration, noise reduction and interpolation to process a received sensing signal, and
wherein the received sensing signal comprises human movement data collected by the at least one data collecting terminal,
wherein the at least one data collecting terminal acquires human movement data from sensors arranged across a human body, and
wherein locations of the sensors are at positions connected with movement of the human body;
the activity recognition system being characterized in that each of the at least one edge computing devices further comprises a model training module and an activity recognition module, wherein:
the model training module retrieves a local activity recognition model by continuously verifying user IDs, and uses the first data to train a versatile network structure and an individualized network structure of the local activity recognition model in a way that individuation features of the user and versatility features of the local activity recognition model are fused with each other, so that the personnel activity recognition conducted by the activity recognition module based on the local activity recognition model obtained after training is balanced between versatility and individuation, and
wherein the local activity recognition model primarily comprises a user identifier and an activity recognizer, the activity recognizer corresponds to the versatile network structure part of the activity recognition model and uses a structured eigenvector as an input to output an activity type, while the user identifier corresponds to the individualized network structure part of the activity recognition model and uses a higher hidden layer in the activity recognizer as an input, wherein model parameters are adjusted continuously with changes in posture features and behavioral habits of a user of the activity recognition system,
and
wherein the user identifier uses a loss function to control it to represent personal features of users, and
wherein each model training module uploads second data it obtains through calculation based on the local activity recognition model before training and the local activity recognition model after training to the cloud computing platform, the cloud computing platform maintains at least one versatile model therein and when the second data uploaded by each said model training module satisfy a predetermined model updating condition, parameters of the at least one versatile model are adjusted, and
wherein each model training module continuously verifies the user IDs so as to obtain new user information or registered user information, when the ID of the current user is the new user information, the at least one versatile model of the cloud computing platform is retrieved and used as the local activity recognition model, or when the ID of the current user is the registered user information, the local activity recognition model in the activity recognition module that corresponds to the current user and has received at least one session of training for activity recognition is retrieved, and
wherein when the ID of the current user is the new user information, the model training module trains the local activity recognition model it retrieves from the cloud computing platform by retaining a fusing result of the versatile network structure and initializing parameters of the individualized network structure, and
wherein the predetermined model updating condition in the cloud computing platform refers to that when a proportion of said local activity recognition models that have finished the model updating exceeds a given threshold, the model updating for the at least one versatile model in the cloud computing platform begins, and
wherein updating of the at least one versatile model is achieved by:
each of said at least one edge computing devices uses a local data calculating model to update a gradient, uses an encryption technique to encrypt the gradient, and sends the encrypted gradient to the cloud computing platform; and/or
the cloud computing platform, without knowledge of any edge computing device information, performs secure aggregation, when the model updating condition for the at least one versatile model is satisfied, on the encrypted gradients it receives, so as to obtain an aggregated gradient; and/or
the cloud computing platform adjusts the at least one versatile model according to the aggregated gradient, thereby achieving updating of the at least one versatile model, and sends the aggregated gradient to each of the at least one edge computing devices; and/or
each of the at least one edge computing devices fine-tunes the respective local activity recognition model according to the aggregated gradient.
|