Deep-learning based peak detection in biological signal

1 . An electronic device, comprising:

circuitry configured to:

receive a biological signal associated with a user;

apply a low-pass filter on the received biological signal;

remove a noise component from the received biological signal based on the application of the low-pass filter on the received biological signal;

determine a denoised signal based on the removal of the noise component from the received biological signal;

detect each of a first peak and a first trough associated with the first peak, from the determined denoised signal;

apply a local search algorithm on the determined denoised signal;

refine each of the detected first peak and the detected first trough based on the application of the local search algorithm on the determined denoised signal; and

determine a health condition associated with the user based on the refined first peak and the refined first trough.

2 . The electronic device according to claim 1 , wherein the low-pass filter corresponds to a moving average filter with a specific window size.

3 . The electronic device according to claim 1 , wherein the circuitry is further configured to:

apply a trained neural network model on the determined denoised signal; and

detect each of the first peak and the first trough based on the application of the trained neural network model on the determined denoised signal.

4 . The electronic device according to claim 3 , wherein the trained neural network model is a scalable deep-learning model that comprises an encoder model, a decoder model, and a set of convolution neural network layers.

5 . The electronic device according to claim 4 , wherein a kernel size associated with each of the set of convolution neural network layers is even.

6 . The electronic device according to claim 3 , wherein

the circuitry is further configured to receive a dataset comprising a set of biological signal recordings that is associated with a set of patients,

the set of patients suffers from the health condition, and

the trained neural network model is trained based on the received dataset.

7 . The electronic device according to claim 1 , wherein the circuitry is further configured to:

transform the received biological signal into a distance transform (DT) map; and

detect each of the first peak and the first trough based on the DT map.

8 . The electronic device according to claim 1 , wherein the local search algorithm corresponds to a local minimum and maximum search algorithm.

9 . The electronic device according to claim 1 , wherein the biological signal corresponds to an electrocardiogram (ECG) of the user.

10 . The electronic device according to claim 9 , wherein

the first peak corresponds to a peak of at least one of a P-wave associated with the ECG, an R-wave associated with the ECG, or a T-wave associated with the ECG, and

the first trough corresponds to a trough of at least one of a Q-wave associated with the ECG, or an S-wave associated with the ECG.

11 . The electronic device according to claim 1 , wherein the biological signal corresponds to an electroencephalogram (EEG) associated with the user.

12 . The electronic device according to claim 11 , wherein the first peak corresponds to an alpha-wave peak associated with the EEG.

13 . The electronic device according to claim 1 , wherein the circuitry is further configured to refine each of the detected first peak and the detected first trough based on baseline information associated with the received biological signal.

14 . The electronic device according to claim 1 , wherein the health condition associated with the user is at least one of a heart condition associated with the user or a brain condition associated with the user.

15 . A method, comprising:

in an electronic device:

receiving a biological signal associated with a user;

applying a low-pass filter on the received biological signal;

removing a noise component from the received biological signal based on the application of the low-pass filter on the received biological signal;

determining a denoised signal based on the removal of the noise component from the received biological signal;

detecting, each of a first peak and a first trough that is associated with the first peak, from the determined denoised signal;

applying a local search algorithm on the determined denoised signal;

refining each of the detected first peak and the detected first trough based on the application of the local search algorithm on the determined denoised signal; and

determining a health condition associated with the user based on the refined first peak and the refined first trough.

16 . The method according to claim 15 , further comprising:

applying a trained neural network model on the determined denoised signal; and

detecting each of the first peak and the first trough based on the application of the trained neural network model on the determined denoised signal.

17 . The method according to claim 15 , further comprising:

transforming the determined denoised signal into a distance transform (DT) map; and

detecting each of the first peak and the first trough based on the DT map.

18 . A non-transitory computer-readable medium having stored thereon, computer-executable instructions that when executed by an electronic device, causes the electronic device to execute operations, the operations comprising:

receiving a biological signal associated with a user;

applying a low-pass filter on the received biological signal;

removing a noise component from the received biological signal based on the application of the low-pass filter on the received biological signal;

determining a denoised signal based on the removal of the noise component from the received biological signal;

detecting each of a first peak and a first trough that is associated with the first peak, from the determined denoised signal;

applying a local search algorithm on the determined denoised signal;

refining each of the detected first peak and the detected first trough based on the application of the local search algorithm on the determined denoised signal; and

determining a health condition associated with the user based on the refined first peak and the refined first trough.