Apparatus and method for diagnosing obstructive sleep apnea

1. A method of diagnosing obstructive sleep apnea (OSA), the method comprising:

acquiring a sleep sound signal comprising sounds made by a person during sleep;

detecting a plurality of snore sounds in the sleep sound signal;

determining a set of mel-frequency cepstral coefficients for each of the snore sounds;

determining a variance for each of the sets of mel-frequency cepstral coefficients;

summing the determined variances;

determining a characterizing feature for the sleep sound signal responsive to the sum of the variances of the cepstral coefficients; and

using the characterizing feature to diagnose OSA in the person, wherein the characterizing feature comprises a numerical value used as an indication of severity of OSA, where a low value means no OSA or a clinically insignificant case of OSA, a medium value means mild OSA, and a high value means severe OSA.

2. A method according to claim 1 and comprising:

determining a plurality of groups of the snore sounds:

determining a group feature for each of the groups;

determining a characterizing feature for the sleep sound signal responsive to the group features; and

using the determined characterizing feature for the sleep sound signal to diagnose OSA in the person.

3. A method according to claim 2 wherein determining a group of snore sounds comprises determining a cluster of consecutive snore sounds in the detected snore sounds for which a time delay between any two temporally adjacent snore sounds is less than or equal to a predetermined time period.

4. A method according to claim 2 wherein the time period is equal to about a minute.

5. A method according to claim 2 wherein determining a group feature for each group comprises determining a measure of energy for each of the snore sounds in the group.

6. A method according to claim 5 wherein determining the group feature comprises using the determined energy measures to determine a measure of an average energy of the snore sounds in the group.

7. A method according to claim 6 wherein determining the group feature comprises using the measure of average energy to determine a variance of the measures of snore sound energies for the group.

8. A method according to claim 7 and comprising determining the characterizing feature of the sleep sound signal responsive to an average of the determined variances of the groups.

9. A method according to claim 7 wherein determining a characterizing feature of the sleep sound signal comprises determining a number of groups in the sound signal for which the variance is greater than a predetermined threshold variance.

10. A method according to claim 1 and comprising:

determining a number of silent periods in the sleep sound signal that are indicative of substantially total suspension of breathing by the person;

determining a characterizing feature of the sleep sound signal responsive to the number of determined silent periods; and

using the characterizing feature to diagnose OSA in the person.

11. A method according to claim 1 and comprising:

determining a pitch density for each of the plurality of snore sounds in the sleep sound signal;

determining an average pitch density for the snore sounds;

determining a characterizing feature of the sleep sound signal responsive to the average pitch density; and

using the characterizing feature to diagnose OSA in the person.

12. A method according to claim 1 wherein using the characterizing sleep sound feature to diagnose OSA comprises providing a figure of merit as to whether the person has OSA that is a linear function of the sleep sound characterizing feature.

13. A method according to claim 12 and comprising configuring the linear function so that the figure of merit is correlated with an apnea-hypopnia index (AHI).

14. A method of diagnosing OSA, the method comprising:

acquiring a sleep sound signal comprising sounds made by a person during sleep;

detecting a plurality of snore sounds in the sleep sound signal;

determining a set of mel-frequency cepstral coefficients for each of the snore sounds;

determining a variance for each of the sets of mel-frequency cepstral coefficients;

summing the determined variances;

determining a plurality of characterizing features for the sleep sound signal, the features comprising:

a first feature determined responsive to the sum of the variances of cepstral coefficients of the snore sounds;

a second feature determined responsive to a measure of an average of variances in energies of snore sounds in groups of the snore sounds;

a third feature determined responsive to a number of groups of snore sounds that have a variance in snore sounds energies greater than a predetermined variance; and

using the determined features to diagnose OSA in the person, wherein the plurality of characterizing features are used to obtain a numerical value used as an indication of severity of OSA, where a low value means no OSA or a clinically insignificant case of OSA, a medium value means mild OSA, and a high value means severe OSA.

15. A method according to claim 14 wherein the indication comprises a figure of merit generated responsive to a linear function of the features.

16. A method according to claim 15 and comprising configuring the linear function so that the figure of merit is correlated with an apnea-hypopnia index (AHI).

17. A method according to claim 14 wherein the plurality of features comprises a fourth feature determined responsive to a number of silent periods in the sleep sound signal that are indicative of substantially total suspension of breathing by the person.

18. A method according to claim 14 wherein the plurality of features comprises a fourth feature determined responsive to an average pitch density for the snore sounds.

19. Apparatus for diagnosing OSA, the apparatus comprising:

a microphone for acquiring a sleep sound signal of a person; and

a computer system configured to:

detect a plurality of snore sounds in the sleep sound signal;

determine a set of mel-frequency cepstral coefficients for each of the snore sounds;

determine a variance for each of the sets of mel-frequency cepstral coefficients;

summing the determined variances;

determine a characterizing feature for the sleep sound signal responsive to the sum of the variances of the cepstral coefficients; and

use the characterizing feature to diagnose OSA in the person, wherein the characterizing feature comprises a numerical value used as an indication of severity of OSA, where a low value means no OSA or a clinically insignificant case of OSA, a medium value means mild OSA, and a high value means severe OSA.