Method for the detecting electrocardiogram anomalies and corresponding system

A device, comprising:

• an interface, which, in operation receives a test heartbeat signal:
• and
• signal processing circuitry, which, in operation:
• determines a heart-rate associated with the test heartbeat signal;
• selects a transform based on the determined heart-rate associated with the test heartbeat signal and a reference heart-rate associated with a dictionary of a sparse approximation model, wherein the transform is selected independent of other factors associated with generation of the test heartbeat signal;
• applies the selected transform to the dictionary of the sparse approximation model, generating an adjusted dictionary of the sparse approximation model;
• and
• detects anomalous heartbeats in the test heartbeat signal using the adjusted dictionary of the sparse approximation model.

A system, comprising:

• a sensor, which, in operation, generates heartbeat signals;
• an accelerometer, which, in operation, generates one or more acceleration signals;
• and signal processing circuitry, which, in operation:
• determines a heart-rate associated with a test heartbeat signal;
• selects a transform based on the determined heart-rate associated with the test heartbeat signal and a reference heart-rate associated with a dictionary of a sparse approximation model, wherein the transform is selected independent of other factors associated with generation of the test heartbeat signal;
• applies the selected transform to the dictionary of the sparse approximation model, generating an adjusted dictionary of the sparse approximation model;
• and
• detects anomalous heartbeats in the test heartbeat signal using the adjusted dictionary of the sparse approximation model.

A non-transitory computer-readable medium having contents which configure a heart-rate monitoring device to perform a method, the method comprising:

• determining a heart-rate associated with a test heartbeat signal;
• selecting a transform based on the determined heart-rate associated with the test heartbeat signal and a reference heart-rate associated with a dictionary of a sparse approximation model, wherein the transform is selected independent of other factors associated with generation of the test heartbeat signal;
• applying the selected transform to the dictionary of the sparse approximation model, generating an adjusted dictionary of the sparse approximation model;
• and
• detecting anomalous heartbeats in the test heartbeat signal using the adjusted dictionary of the sparse approximation model.

A method, comprising:

• learning a plurality of dictionaries of sparse representation models of heartbeats associated with a plurality of users and a plurality of heart-rates;
• solving an optimization problem for multiple sets of dictionaries of the plurality of dictionaries with respect to pairs of the plurality of heart-rates, generating a family of transforms;
• selecting a transform of the family of transforms based on a target heart rate and a reference heart-rate, wherein the reference heart-rate is associated with a dictionary of another sparse representation model of heartbeat;
• and
• applying the selected transform to the dictionary of the another sparse representation model, generating an adjusted dictionary.

A non-transitory computer-readable medium having contents which cause processing circuitry to perform a method, the method comprising:

• learning a plurality of dictionaries of sparse representation models of heartbeats associated with a plurality of users and a plurality of heart-rates;
• solving an optimization problem for multiple sets of dictionaries of the plurality of dictionaries with respect to pairs of the plurality of heart-rates, generating a family of transforms;
• selecting a transform of the family of transforms based on a target heart rate and a reference heart-rate, wherein the reference heart-rate is associated with a dictionary of another sparse representation model of heartbeat;
• and
• applying the selected transform to the dictionary of the another sparse representation model, generating an adjusted dictionary.

A system, comprising:

• one or more memories;
• and
• processing circuitry, which, in operation:
• learns a plurality of dictionaries of sparse representation models of heartbeats associated with a plurality of users and a plurality of heart-rates;
• solves an optimization problem for multiple sets of dictionaries of the plurality of dictionaries with respect to pairs of the plurality of heart-rates, generating a family of transforms;
• selects a transform of the family of transforms based on a target heart rate and a reference heart-rate, wherein the reference heart-rate is associated with a dictionary of another sparse representation model of heartbeat;
• and
• applies the selected transform to the dictionary of the another sparse representation model, generating an adjusted dictionary.