Method and system for activity classification
Inventors
Cheng, Chun Hing • EVERETT, Julia Breanne • PURDY, MICHAEL TODD • STEVENS, Travis Michael • VIBERG, DAVID ALLAN • YEE, DALE BARRY
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-11853887-B2
Publication Date
2023-12-26
Expiration Date
Abstract
A method and system for activity classification. A pressure sensor receives input data resulting from physical activity of a subject performing an activity. The input data includes pressure data from at least one pressure sensor, and may include other data acquired through other types of sensors. A deep learning neural network is applied to the input data for identifying the activity. The neural network is trained with reference to training data from a training database. The training data may include empirical data from a database of previous data of corresponding activities, synthesized data prepared from the empirical data or simulated data. The training data may include data from physical activity of the subject being monitored by the system. Different aspects of the neural network may be trained with reference to the training data, and some aspects may be locked or opened depending on the application and the circumstances.
Core Innovation
The invention determines energy expenditure data of a subject by generating a first set of input data during an activity, where the first set of input data comprises pressure data. A first deep learning neural network is applied to the first set of input data based on a first set of weights and biases to produce classified activity data, and the first deep learning neural network is trained for updating the first set of weights and biases.
A second deep learning neural network is then applied to a second set of input data comprising the classified activity data, based on a second set of weights and biases, resulting in energy expenditure data. The invention can further segment input over time using time window processing, producing time-segmented input data for the first deep learning neural network.
Training can be performed using confirmation input, a loss function, and updating weights and biases based on classified actual activity data and classified activity data. The invention can also enrich the first and/or second input sets with personal and/or physical attributes of the subject, including health information obtained from a database and/or self-inputted health information, and can include data obtained from a biological feature sensor in the second set of input data.
The energy expenditure data can be communicated to a drug requirement calculator to generate drug requirement data and to drive drug delivery using a drug delivery system. The document also describes insulin-related embodiments in which the drug requirement data relate to insulin, the drug delivery system is an insulin delivery system, and the drug is insulin.
Claims Coverage
The provided claim set includes two independent claims, a method claim and a system claim. Across these, the core inventive features comprise generating pressure-based input data, applying a first deep learning neural network to produce classified activity data, training that network by updating weights and biases, and applying a second deep learning neural network to generate energy expenditure data from the classified activity data.
Pressure-data activity input to classified activity data via first deep learning neural network
Generating a first set of input data of the subject resulting from an activity, the first set of input data comprising pressure data; applying a first deep learning neural network to the first set of input data based on a first set of weights and biases, resulting in classified activity data.
Training first deep learning neural network to update weights and biases
Training the first deep learning neural network for updating the first set of weights and biases.
Energy expenditure data generation via second deep learning neural network
Applying a second deep learning neural network to a second set of input data comprising the classified activity data based on a second set of weights and biases, resulting in energy expenditure data.
Pressure-sensor module system producing energy expenditure data
A sensor module comprising a pressure sensor for generating a first set of input data during an activity, the first set of input data comprising pressure data; a processor included in a first device configured for receiving the first set of input data and executing applying a first deep learning neural network resulting in classified activity data, training the first deep learning neural network for updating the first set of weights and biases, and applying a second deep learning neural network to the second set of input data comprising the classified activity data resulting in energy expenditure data.
Downstream drug requirement calculation from communicated energy expenditure data
Communicating energy expenditure data to a drug requirement calculator to generate drug requirement data for the subject.
Both independent claims cover a pressure-sensor/pressure-data input pipeline that uses a first deep learning neural network to generate classified activity data, with weights-and-biases training, and a second deep learning neural network to convert the classified activity data into energy expenditure data. The system claim additionally specifies a sensor module and device processor implementation, and the claims also include cascading to drug requirement data via communication to a drug requirement calculator.
Stated Advantages
Not explicitly described in patent.
Documented Applications
Not explicitly described in patent.
Interested in licensing this patent?