Methods and apparatus for performing dynamic respiratory classification and tracking

1. A method for performing acoustic dynamic classification of a breathing session using a processor coupled to a memory, said method comprising:

capturing breathing sounds of a subject using a microphone;

processing said breathing sounds to generate an audio respiratory signal;

recognizing a plurality of breath cycles and a plurality of breath phases within each of said plurality of breath cycles from said audio respiratory signal;

detecting characteristics regarding said plurality of breath cycles and said plurality of breath phases; and

extracting and outputting metrics concerning a breath pattern quality of said subject using said characteristics wherein said metrics are selected from a group consisting of: respiratory rate; depth; tension; nasal wheeze; tracheal wheeze; pre-apnea; apnea; ramp; flow; variability; and inhale/exhale ratio;

wherein said recognizing comprises:

obtaining a first audio envelope by filtering an audio respiratory signal, wherein said obtaining comprises calculating a spectral centroid of each block of said audio respiratory signal and filtering said audio respiratory signal with a low pass filter tuned to a minimum value of said spectral centroid;

classifying lobes of said first audio envelope into a plurality of classes; and

defining said plurality of breath cycles and said plurality of breath phases using timestamps obtained from said classifying.

2. The method of claim 1 , wherein each breath phase within said plurality of breath phases is selected from a group consisting of: inhale; transition; exhale and rest.

3. The method of claim 1 , wherein each class within said plurality of classes is selected from a group consisting of: inhalations; exhalations; rest periods.

4. The method of claim 1 , wherein said classifying comprises:

calculating a second audio envelope using a window, wherein said window has a size set according to a respiratory rate;

normalizing said second audio envelope;

detecting peaks of said second audio envelope;

determining a start timestamp and an end timestamp of each breath cycle using said peaks; and

calculating a low threshold and a high threshold by using a moving average filter on said second audio envelope, wherein said low threshold detects signal presence, and further wherein said high threshold discriminates between inhalation and exhalation events.

5. The method of claim 4 , wherein calculating said respiratory rate comprises:

calculating an auto-correlation function (ACF) of said first audio envelope; and

determining a peak of said ACF; and

using said peak to determine said respiratory rate.

6. The method of claim 5 , wherein said respiratory rate is used to determine a ventilatory threshold wherein said ventilatory threshold occurs when a rate of change of said respiratory rate is greatest.

7. The method of claim 1 , wherein said characteristics are selected from a group consisting of: inhale duration, pause duration, exhale duration, rest duration, a wheeze source, a wheeze type, a cough type, a cough source, choppiness, smoothness, attack and decay.

8. The method of claim 1 , wherein at least one of said characteristics is wheezing, and wherein said detecting comprises:

analyzing a block of said audio respiratory signal;

calculating an auto-correlation function (ACF) for said block;

calculate a linear predictive coefficient (LPC) for said block;

generating an inverse LPC filter along with a magnitude response for each block; and

analyzing said ACF and said magnitude response to identify said wheezing.

9. The method of claim 1 , wherein at least one characteristic is coughing, and wherein said detecting comprises:

extracting a first plurality of descriptors that define a cough pattern from said audio respiratory signal;

comparing said first plurality of descriptors to a database comprising a second plurality of descriptors extracted from sample coughs; and

mapping an input cough from said audio respiratory signal to a closest match from said database.

10. The method of claim 9 , wherein said descriptors are selected from a group consisting of: attack time; decay time; envelope intensity; spectral centroid; spectral spread; spectral kyrtosis; and harmonicity.

11. The method of claim 10 , wherein said descriptors are used to perform a spirometry analysis.

12. The method of claim 1 , wherein said descriptors can be used to determine a respiratory compensation threshold (RCT).

13. A computer-readable storage medium having stored thereon, computer executable instructions that, if executed by a computer system cause the computer system to perform a method for performing acoustic dynamic classification of a breathing session using a processor coupled to a memory, said method comprising:

capturing breathing sounds of a subject using a microphone;

processing said breathing sounds to generate an audio respiratory signal;

recognizing a plurality of breath cycles and a plurality of breath phases within each of said plurality of breath cycles from said audio respiratory signal;

detecting characteristics regarding said plurality of breath cycles and said plurality of breath phases; and

extracting and outputting metrics concerning a breath pattern quality of said subject using said characteristics wherein said metrics are selected from a group consisting of: respiratory rate; depth; tension; nasal wheeze; tracheal wheeze; pre-apnea; apnea; ramp; flow; variability; and inhale/exhale ratio;

wherein said recognizing comprises:

obtaining a first audio envelope by filtering an audio respiratory signal, wherein said obtaining comprises calculating a spectral centroid of each block of said audio respiratory signal and filtering said audio respiratory signal with a low pass filter tuned to a minimum value of said spectral centroid;

classifying lobes of said first audio envelope into a plurality of classes; and

defining said plurality of breath cycles and said plurality of breath phases using timestamps obtained from said classifying.

14. The computer-readable medium as described in claim 13 , wherein said classifying comprises:

calculating a second audio envelope using a window, wherein said window has a size set according to a respiratory rate;

normalizing said second audio envelope;

detecting peaks of said second audio envelope;

determining a start timestamp and an end timestamp of each breath cycle using said peaks; and

calculating a low threshold and a high threshold by using a moving average filter on said second audio envelope, wherein said low threshold detects signal presence, and further wherein said high threshold discriminates between inhalation and exhalation events.

15. The computer-readable medium as described in claim 14 , wherein calculating said respiratory rate comprises:

calculating an auto-correlation function (ACF) of said first audio envelope; and

determining a peak of said ACF; and

using said peak to determine said respiratory rate.

16. The computer-readable medium as described in claim 13 , wherein at least one of said characteristics is wheezing, and wherein said detecting comprises:

analyzing a block of said audio respiratory signal;

calculating an auto-correlation function (ACF) for said block;

calculating a linear predictive coefficient (LPC) for said block;

generating an inverse LPC filter along with a magnitude response for each block; and

analyzing said ACF and said magnitude response to identify said wheezing.

17. An apparatus for performing acoustic dynamic classification of a breathing session, said apparatus comprising:

a microphone for capturing breathing sounds of a subject;

a memory comprises an application for performing dynamic classification of a breathing session stored therein; and

a processor coupled to said memory and said microphone, the processor being configured to operate in accordance with said application to:

process said breathing sounds to generate an audio respiratory signal;

recognize a plurality of breath cycles and a plurality of breath phases within each of said plurality of breath cycles from said audio respiratory signal;

detect characteristics regarding said plurality of breath cycles and said plurality of breath phases; and

extract and output metrics concerning a breath pattern quality of said subject using said characteristics wherein said metrics are selected from a group consisting of: respiratory rate; depth; tension; nasal wheeze; tracheal wheeze; pre-apnea; apnea; ramp; flow; variability; and inhale/exhale ratio;

wherein to recognize said plurality of breath cycles and said plurality of breath phases, said processor is configured to:

obtain a first audio envelope from said audio respiratory signal by calculating a spectral centroid of each block of said audio respiratory signal and filtering said audio respiratory signal with a low pass filter tuned to a minimum value of said spectral centroid;

classify lobes of said first audio envelope into a plurality of classes; and

define said plurality of breath cycles and said plurality of breath phases using timestamps obtained from said classifying.