System

A system comprising:

• circuitry configured to:
• receive, via a communication interface operating over a packet-switched network using a transport layer security protocol, time-series data sets from a terminal device, the time-series data sets representing preprocessed digital bioelectrical signals acquired via a plurality of electrodes attached to a body surface of a user and sampled at a predetermined sampling frequency, each time-series data set corresponding to a signal interval of a predetermined time length and comprising a two-dimensional array of normalized sample values indexed by signal channels and temporal indices together with metadata including a time stamp and a user identifier;
• store the time-series data sets in a storage medium indexed by the user identifier and the time stamp, and convert the time-series data sets into tensor data by arranging the two-dimensional arrays into a multi-dimensional numerical array structure and stacking a plurality of consecutive data sets to form a three-dimensional tensor having axes corresponding to signal channels, time, and a data set index;
• execute inference processing by inputting the tensor data to a trained information processing model comprising a convolution computation layer configured to extract spatial and frequency features by applying learnable filters having kernel sizes spanning a plurality of signal channels and a temporal window with an activation function and a batch normalization operation, a pooling operation to reduce temporal resolution, and a recurrent computation layer comprising a long short-term memory network configured to update a hidden state at each time step based on pooled feature vectors, and a fully connected output layer with a softmax function to compute probabilities for each of a plurality of predefined intention classes;
• select an intention class having a highest probability as intention information subject to a confidence threshold, and generate high-level command data comprising at least an action type, a target parameter, an intensity value, a time stamp, and a configuration flag specifying a maximum allowable output value based on the intention information;
• transmit the high-level command data to the terminal device via the packet-switched network using a persistent messaging connection;
• record, in the storage medium, the time-series data sets, the intention information, the high-level command data, and execution result information received from the terminal device in association with one another, the execution result information indicating at least one of a successful completion, an interruption, and an error condition;
• perform adaptive learning processing comprising partitioning the recorded information into training and validation data sets, computing a loss function measuring a discrepancy between predicted intention probabilities and reference labels, computing gradients using backpropagation, and updating parameters of the trained information processing model using an adaptive gradient-based optimization algorithm;
• and
• execute a generative design process in which a generative artificial intelligence model receives at least one prompt sentence describing a structure of the trained information processing model, signal-processing conditions, and a format of the high-level command data, and automatically generates or refines at least a part of the trained information processing model based on the prompt sentence.

A method performed by circuitry of a system, the method comprising:

• receiving, via a communication interface operating over a packet-switched network, time-series data sets from a terminal device, the time-series data sets representing preprocessed digital signals acquired from a plurality of sensing elements;
• converting the time-series data sets into tensor data and executing inference processing using a trained information processing model that comprises a convolution computation layer configured to extract spatial and frequency features and a recurrent computation layer configured to analyze temporal transitions of the features, so as to output intention information corresponding to a predetermined action of a user;
• generating high-level command data comprising at least an action type, a target parameter, and an intensity value based on the intention information, and transmitting the high-level command data to the terminal device via the packet-switched network;
• recording, in a storage medium, the time-series data sets, the intention information, the high-level command data, and execution result information received from the terminal device in association with one another;
• and
• performing adaptive learning processing to update parameters of the trained information processing model on a per-user basis using the recorded information.