SOUND SOURCE LOCALIZATION AND ISOLATION APPARATUSES, METHODS AND SYSTEMS

• US 20150156578A1
• Filed: 11/28/2014
• Published: 06/04/2015
• Est. Priority Date: 09/26/2012
• Status: Active Grant

• Current Assignee: Foundation For Research & Technology
• Original Assignee: Foundation For Research & Technology
• Inventors: Alexandridis, Anastasios, Griffin, Anthony, Mouchtaris, Athanasios
• Granted Patent: US 9,549,253 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: H04R 2420/07 Applications of wireless loudspeakers or wireless microphones
• CPC Class Codes: H04R 2430/03 Synergistic effects of band splitting and sub-band processing
• CPC Class Codes: H04R 3/005 for combining the signals of two or more microphones specially adapted for hearing aids H04R25/407
• CPC Class Codes: H04S 2420/01 Enhancing the perception of the sound image or of the spatial distribution using head related transfer functions [HRTF's] or equivalents thereof, e.g. interaural time difference [ITD] or interaural level difference [ILD]

Abstract

A processor-implemented method for spatial sound localization and isolation is described. The method includes segmenting, via a processor, each of a plurality of source signals detected by a plurality of sensors, into a plurality of time frames. For each time frame, the method further includes obtaining, via a processor, a plurality of direction of arrival (DOA) estimates from the plurality of sensors, discretizing an area of interest into a plurality of grid points, calculating, via the processor, DOA at each of grid points, comparing, via the processor, the DOA estimates with the computed DOAs. If the number of sources is more than 1, the method includes obtaining via the processor, a plurality of combinations of DOA estimates, from amongst the plurality of combinations, estimating, via the processor, one or more initial candidate locations corresponding to each of the combinations, selecting location of the sources from amongst the initial candidate locations.

Claims

1. A processor-implemented method for spatial sound localization, the method comprising:

obtaining, via a processor, a plurality of direction of arrival (DOA) estimates from a plurality of sensors;

determining, via the processor, a set of intersection points based on the DOA estimates;

receiving, via the processor, a number of sources in a current time frame;

if the number of sources is more than 1, obtaining via the processor, a plurality of regions by dividing a plurality of possible locations of sources into a predefined number of combinations of DOA estimates;

from amongst the plurality of regions, selecting, via the processor, a region having maximum number of intersection points;

obtaining via the processor, a centroid of the intersection points in the selected region;

estimating location of one of the plurality of sources based on the centroid;

selecting the remaining region and obtaining a centroid of intersection points in the remaining region to yield location of the remaining source.

2. The method of claim 1, wherein the region is selected from amongst the plurality of regions based on a minimum value of variance.

3. The method of claim 1, wherein the possible locations are based at least on the intersection points of a pair of DOA estimates that are not substantially parallel, and wherein the pair of DOA estimates belong to distinct sensors.

4. The method of claim 1 further comprising:

determining one or more intersection point outliers based at least on a parallelness threshold; and

removing the intersection point outliers from the set of intersection points to yield an updated set of intersection points.

5. The method of claim 4 further comprising:

determining a centroid of the intersection points remaining after removing the intersection point outliers, if the number of sources is equal to one; and

estimating location of the source based on the centroid.

6. The method of claim 4, wherein an angular distance between the plurality of sensors is compared with the parallelness threshold.

7. A system for spatial sound localization, the system comprising:

a processor;

a memory coupled to the processor, the memory comprising,

a sound source localizer configured to,

obtain a plurality of direction of arrival (DOA) estimates from a plurality of sensors;

determine a set of intersection points based on the DOA estimates;

determine one or more intersection point outliers based at least on a parallelness threshold;

remove the intersection point outliers from the set of intersection points to yield an updated set of intersection points;

determine a centroid of the intersection points remaining after removing the intersection point outliers; and

estimate location of one or more sources based on the centroid.

8. The system of claim 7, wherein the sound source localizer includes a region determination module configured to:

determine via the processor, a plurality of regions by dividing a plurality for each of the plurality of sources into a predefined number of unique combination of the DOA estimates, wherein the possible locations are based at least on the intersection points of the pair of estimates that are not substantially parallel;

arrange the plurality of regions in a list according to descending number of intersection points;

from amongst the plurality of regions, select, via the processor, a region having maximum number of intersection points;

obtain via the processor, a centroid of the intersection points in the selected region;

estimate location of one of the plurality of sources based on the centroid;

select another region from the list; and

obtain a centroid of intersection points in the another region to yield location of the remaining source.

9. A method for sound source localization, the method comprising:

segmenting, via a processor, each of a plurality of source signals detected by a plurality of sensors, into a plurality of time frames;

for each time frame,

obtaining, via a processor, a plurality of direction of arrival (DOA) estimates from the plurality of sensors;

discretizing an area of interest into a plurality of grid points;

calculating, via the processor, DOA at each of grid points;

comparing, via the processor, the DOA estimates with the computed DOAs;

if the number of sources is more than 1,

obtaining via the processor, a plurality of combinations of DOA estimates;

from amongst the plurality of combinations, estimating, via the processor, one or more initial candidate locations corresponding to each of the combinations;

selecting location of the sources from amongst the initial candidate locations.

10. The method of claim 9, wherein selecting location of the sources includes searching a plurality of S-tuples of DOA combinations.

11. The method of claim 9, wherein selecting location of the sources includes:

determining a residual value based on a relationship between the DOA estimates and DOA estimates at the initial candidate location; and

evaluating DOA combinations that approximate the residual.

12. The method of claim 9, wherein obtaining DOA estimates includes:

obtaining DOA estimates for each frequency of at least a current time frame.

and a plurality of a predefined number of previous frames.

13. The method of claim 9, wherein comparing includes:

generating histograms using the frequency distribution at a plurality of time frames;

comparing histograms based on a correlation coefficient; and

associating a DOA estimate with a source based on the comparison.

14. The method of claim 13 further comprising removing the associated DOA and determining association of another DOA estimate with another source.

15. The method of claim 9 further comprising:

determining whether the number of sources is equal to one;

if the number of sources is equal to one, selecting a grid point, wherein computed DOA at the selected grid point matches with its DOA estimate; and

estimating location of the source based on the grid point.

16. The method of claim 9, wherein angular distance between the plurality of sensors is compared to determine a match.

17. The method of claim 9, wherein selecting a grid point further comprises:

receiving a resolution coefficient;

determining a updated grid point based on the resolution coefficient, wherein the updated grid point is centered on the grid point.

18. A processor-implemented method for sound source isolation, the method comprising:

deriving, via the processor, time-frequency transform of a plurality of source signals;

obtaining an estimated number of sources and their respective DOAs; and

determining, via the processor, an array from amongst the plurality of arrays based at least on proximity to a source;

determining DOA corresponding to the microphone array;

extracting spatially separated source signals based at least on the estimated DOA of the source in proximity to the array; and

processing, via the processor, spatially separated source signals to yield at least one reference signal and side information.

19. The method of claim 18 further comprising filtering the spatially separated source signals based at least on binary masks, wherein the number of binary masks varies based on the estimated number of sources.

20. The method of claim 18, wherein a single post-filtered signal is obtained based on the spatially separated source signals.

21. The method of claim 20, wherein the number of beamformers is calculated based on the estimated number of sources.

22. The method of claim 18 further comprises encoding the side information, wherein the encoding includes:

arranging the estimated sources according to an assigned number of frequency bins;

based on orthogonality, encoding one or more binary masks of each sound source; and

inserting the binary masks in a bitstream as encoded side information.

23. The method of claim 18, wherein estimating the number of sound sources includes:

detecting one or more single-source analysis zones based at least on a correlation between the time-frequency transform of the source signals;

estimating at least one direction of arrival for each source in the detected single source analysis zones;

creating a histogram of the estimated directions of arrival from a plurality of sources; and

using a matching pursuit method to estimate the number of sources.

24. An apparatus for localizing a plurality of sound sources using direction of arrival (DOA) estimates from a plurality of sensors, the apparatus comprising:

a memory;

a network;

a processor in communication with the memory and the network, and configured to issue a plurality of processing instructions stored in the memory, wherein the processor issues instructions to;

(a) obtain, via at least one of the network and the memory, the DOA estimates from the plurality of sensors;

(b) form pairs of DOA estimates, wherein each pair includes DOA estimates from distinct sensors;

(c) for each pair of the DOA estimates,

determine, by the processor, whether the pair of DOA estimates are substantially parallel;

discard, by the processor, the pair of DOA estimates if they are substantially parallel; and

determine, by the processor, the intersection point of the pair of DOA estimates if they are not substantially parallel;

(d) determine, by the processor, a plurality of regions by dividing possible locations for each of the plurality of sources into a predefined number of unique combination of the DOA estimates, wherein the possible locations are based at least on the intersection points of the pair of estimates that are not substantially parallel;

(e) select, by the processor, a first one of the plurality of regions containing the most intersection points;

(f) determine, by the processor, the centroid of the intersection points in the selected region, wherein a location of one of the plurality of sound sources is given by the centroid;

(g) select, by the processor, a next one of the plurality of regions containing the most intersection points;

(h) determine, by the processor, the centroid of the intersection points in the next selected region, wherein a location of a next one of the plurality of sound sources is given by the centroid; and

(i) repeat g) and h) until each of the plurality of sound sources have been located.

25. An apparatus for localizing a plurality of sound sources using direction of arrival (DOA) estimates from a plurality of sensors, the apparatus comprising:

a memory;

a network;

a processor in communication with the memory and the network, and configured to issue a plurality of processing instructions stored in the memory, wherein the processor issues instructions to;

(a) obtain, via at least one of the network and the memory, the DOA estimates from the plurality of sensors;

(b) form, by the processor, a set of all possible unique combinations of DOA estimates;

(c) for each unique combination of DOA estimates, determine, by the processor, a plurality of initial candidate locations of the sound sources by,

discretizing an area of interest into a plurality of grid points; and

determining, by the processor, a grid point whose DOA most closely matches the DOA estimates, wherein each determined grid point corresponds to an initial candidate location of a sound source;

(d) determine, by the processor, a correct association of DOA estimates that correspond to the same sound source; and

(e) determine, by the processor, which of the plurality of initial candidate locations most likely correspond to true sound source locations based on the correct association of DOA estimates.