Audio-visual object localization and tracking system and method therefor

• US 6,005,610 A
• Filed: 01/23/1998
• Issued: 12/21/1999
• Est. Priority Date: 01/23/1998
• Status: Expired due to Term

• Current Assignee: RPX Corporation
• Original Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Inventors: Pingali, Sarma VGK
• Primary Examiner(s): Le, Vu
• Application Number:
  US09/012,393
• Time in Patent Office: 697 Days
• Field of Search: 348/169, 348/170, 348/142-143, 348/152-155, 348/162-164, 348/13-15, 348/20, 348/25, 382/103
• US Class Current: 348/169
• CPC Class Codes: G01S 3/7864 T.V. type tracking systems
• CPC Class Codes: G01S 3/808 using transducers spaced apart and measuring phase or time difference between signals therefrom, i.e. path-difference systems

Abstract

A method for integrated audio-visual localizing and tracking of at least one object. The method includes the steps of capturing and transmitting an image of a video scene using a camera at an instant of time, identifying an object contained in the image having a preselected visual feature, and estimating a location of the object by determining an angular orientation relative to the image plane of the camera of an imaginary line extending from an optical center of the camera to a point on the image plane of the camera representing a portion of the object. The method further includes the steps of converting acoustic waves from an audio source into audio signals using at least two microphones at substantially the same time, identifying the audio source by determining a locus of points representing an estimate of the location of the audio source on the basis of the audio signals. An improved estimate of the location of the object is computed by determining the location of a region of intersection between the imaginary line and the locus.

Claims

1. An integrated audio-visual method for localizing and tracking at least one object, comprising the steps of:

capturing and transmitting an image of a video scene using a camera at an instant of time;

identifying an object contained in said image having a preselected visual feature;

estimating a location of the object by determining an angular orientation relative to the image plane of the camera of an imaginary line extending from an optical center of the camera to a point on the image plane of the camera representing a portion of the object;

converting acoustic waves from an audio source into audio signals using at least two microphones at substantially said instant of time;

identifying the audio source by determining on the basis of the audio signals a locus of points representing an estimate of the location of the audio source; and

computing the location of a region of intersection between the imaginary line and the locus, the intersection region being an improved estimate of the location of the object.

2. The integrated audio-visual method of claim 1, further comprising the steps of segmenting a portion of the image containing the identified object;

associating the audio signals with the identified object, the improved location estimate and said instant of time;

associating the portion of the image containing the identified object with the improved location estimate and said instant of time; and

recording the improved location estimate, the segmented portion of the image containing the identified object, the audio signals, and said instant of time.

3. An apparatus for integrated audio-visual localizing and tracking of at least one object:

a camera for capturing an image of a video scene and generating video signals representing the image at an instant of time;

visual object localizer means connected to said camera for receiving the video signals, identifying an object having a preselected visual cue contained in said image, and for estimating a location of the object by determining the angular orientation of an imaginary line extending from an optical center of the camera to a point on an image plane of the camera representing a portion of the object;

at least two microphones operative to pick up and convert acoustic waves from an audio source into audio signals at substantially said instant of time;

audio source localizer means connected to said at least two microphones for determining a locus of points representing an estimate of the location of the audio source generating the acoustical waves on the basis of the audio signals; and

integrated localizer means operatively connected to said visual object localizer means and said audio source localizer means for receiving information relating to said imaginary line and said locus for computing an improved estimate of the location of the object by determining the location of a region of intersection between said imaginary line and said locus of points.

4. The apparatus of claim 3, further comprising another pair of microphones so arranged that imaginary straight lines connecting each pair of microphones intersect perpendicularly with each other so that the locus is a straight line extending from the intersecting point of said imaginary straight lines connecting said microphone pairs.

5. The apparatus of claim 3, wherein said integrated localizer means outputs signals relating to said improved estimate of the location of the object, and further comprising a servomechanism connected to said integrated localizer means and a plurality of cameras and directional microphones connected to said servomechanism, said servomechanism being responsive to the signals from said integrated localizer means to direct said plurality of cameras and directional microphones to the improved estimate of the location of the object computed by said integrated localizer means so as to improve the quality of the image and the pick-up of acoustic waves from the improved estimate of the location of the object.

6. An integrated audio-visual method for localizing and tracking multiple objects, comprising the steps of:

capturing and transmitting images of a video scene using a plurality of cameras at an instant of time;

identifying objects having preselected visual features contained in said images;

determining an angular orientation of each of a plurality of imaginary lines extending from an optical center of each of the plurality of cameras to each point on an image plane of the camera representing a portion of each of the identified objects;

grouping points of intersection of said imaginary lines according to a distance measure;

converting acoustic waves from a plurality of audio sources into audio signals using a plurality of microphones at substantially said instant of time;

determining a plurality of loci of points representing estimates of locations of the audio sources on the basis of the audio signals from pairs of microphones selected from said plurality of microphones; and

determining a region of intersection between each of said plurality of loci and each of said grouped points of intersection of said imaginary lines according to another distance measure, said region of intersection being an improved estimate of a location of an identified object.

7. The integrated audio-visual method of claim 6, further comprising the step of defining each group of points of intersection as a visual cluster, wherein said distance measure employs a root-mean-squared sum of positions of points of intersection of said imaginary lines weighted by confidence factors so that there is an equal number of visual clusters and identified objects.

8. The integrated audio-visual method of claim 6, further comprising the step of defining said region of intersection as an audio-visual cluster, wherein said another distance measure employs a root-mean-squared sum of positions of points of intersection of said plurality of loci and said grouped points of intersection of said imaginary lines weighted by confidence factors so that there is an equal number of audio-visual clusters and identified objects.

9. The method of claim 6, wherein the objects are persons, and further comprising the steps of:

defining each of the grouped points of intersection as a visual cluster;

defining each of the plurality of loci as an audio cluster;

defining said region of intersection as an audio-visual cluster;

associating audio signals used for determining at least a portion of each said audio-visual cluster with an identified person whose location corresponds to the location represented by said audio-visual cluster; and

associating an image used for determining at least a portion of said audio-visual cluster with an identified person whose location corresponds to the location represented by said audio-visual cluster.

10. The method of claim 9, further comprising the steps of:

defining a visual cluster disposed outside said region of intersection as a visual-only cluster; and

associating at least a portion of an image used for determining at least a portion of said visual-only cluster with an identified person whose location corresponds to a location represented by said each said visual-only cluster.

11. The method of claim 10, further comprising the step of recording said associated image of said identified person, said location of said identified person, and said instant of time.

12. The method of claim 9, further comprising the steps of:

defining an audio-only cluster as an audio cluster disposed outside said region of intersection; and

associating audio signals used for determining at least a portion of said audio-only cluster with an audio source whose location corresponds to the location represented by said audio-only cluster.

13. The method of claim 12, further comprising the step of recording said associated audio signals of said audio source, said location of said audio source, and said instant of time.

14. The method of claim 9, further comprising the step of recording said associated image and associated audio signal of said identified person, said location of said identified person, and said instant of time.

15. The method of claim 9, wherein said distance measure employs a root-mean-squared sum of the positions of the points of intersection weighted by confidence factors so that there is an equal number of visual clusters and identified persons.

16. The method of claim 9, wherein said another distance measure employs a root-mean-squared sum of positions of points of intersection of said plurality of loci and said grouped points of intersection of said imaginary lines weighted by confidence factors so that there is an equal number of audio-visual clusters and identified persons.