View source for 18509722. Method for Detecting and Classifying Coughs or Other Non-Semantic Sounds Using Audio Feature Set Learned from Speech simplified abstract (Google LLC)

=Method for Detecting and Classifying Coughs or Other Non-Semantic Sounds Using Audio Feature Set Learned from Speech=


==Organization Name==

[[:Category:Google LLC|Google LLC]]


[[Category:Google LLC]]

==Inventor(s)==

[[:Category:Jacob Garrison of Seattle WA (US)|Jacob Garrison of Seattle WA (US)]][[Category:Jacob Garrison of Seattle WA (US)]]

[[:Category:Jacob Scott Peplinski of Chandler AZ (US)|Jacob Scott Peplinski of Chandler AZ (US)]][[Category:Jacob Scott Peplinski of Chandler AZ (US)]]

[[:Category:Joel Shor of Tokyo (JP)|Joel Shor of Tokyo (JP)]][[Category:Joel Shor of Tokyo (JP)]]

==Method for Detecting and Classifying Coughs or Other Non-Semantic Sounds Using Audio Feature Set Learned from Speech - A simplified explanation of the abstract==


This abstract first appeared for US patent application 18509722 titled 'Method for Detecting and Classifying Coughs or Other Non-Semantic Sounds Using Audio Feature Set Learned from Speech

==Simplified Explanation==

The patent application describes a method for detecting coughs in an audio stream using a self-supervised triplet loss embedding model.

* Pre-processing steps are performed on the audio stream to create an input audio sequence with time-separated audio segments.
* An embedding is generated for each segment using an audio feature set and a self-supervised triplet loss embedding model.
* The model trained on speech audio clips from a dataset provides the embeddings to a cough detection model for inference.
* The cough detection model assigns a probability to each segment containing a cough episode, generating cough metrics for detected episodes.

== Potential Applications ==
This technology can be applied in healthcare settings for remote monitoring of patients with respiratory conditions, in smart home devices for detecting coughing episodes, and in public spaces for monitoring the spread of contagious diseases.

== Problems Solved ==
This technology solves the problem of accurately detecting coughing episodes in audio streams, enabling early intervention for individuals with respiratory issues and providing valuable data for public health monitoring.

== Benefits ==
The benefits of this technology include improved remote patient monitoring, early detection of respiratory issues, efficient public health surveillance, and enhanced data collection for research purposes.

== Potential Commercial Applications ==
* Healthcare monitoring devices
* Smart home systems
* Public health surveillance tools

== Possible Prior Art ==
There may be prior art related to audio-based cough detection systems using machine learning models, but the specific approach of using a self-supervised triplet loss embedding model for cough detection may be novel.

=== Unanswered Questions ===

== How does this technology handle background noise in the audio stream during cough detection? ==
The patent application does not provide details on how the system distinguishes coughing episodes from other sounds or background noise.

== What is the computational efficiency of the cough detection model in real-time applications? ==
The patent application does not discuss the computational requirements or real-time performance of the cough detection model.


==Original Abstract Submitted==

A method of detecting a cough in an audio stream includes a step of performing one or more pre-processing steps on the audio stream to generate an input audio sequence comprising a plurality of time-separated audio segments. An embedding is generated by a self-supervised triplet loss embedding model for each of the segments of the input audio sequence using an audio feature set, the embedding model having been trained to learn the audio feature set in a self-supervised triplet loss manner from a plurality of speech audio clips from a speech dataset. The embedding for each of the segments is provided to a model performing cough detection inference. This model generates a probability that each of the segments of the input audio sequence includes a cough episode. The method includes generating cough metrics for each of the cough episodes detected in the input audio sequence.



[[Category:G10L25/66]]
[[Category:A61B5/00]]
[[Category:A61B5/08]]
[[Category:G10L15/02]]
[[Category:G10L15/04]]
[[Category:G10L15/06]]
[[Category:G10L25/30]]
[[Category:G10L25/51]]
[[Category:G10L25/78]]
[[Category:G16H40/67]]