 |
Claims  |
|
|
What is claimed is:
1. A method for separating at least one sound source from a plurality of sound sources using a plurality of microphones disposed separately from one another, comprising steps
of:
(a) dividing an output channel signal from each microphone into a plurality of frequency bands to produce band-divided output channel signals;
(b) detecting, for each frequency band, as band-dependent inter-channel parameter value differences, differences between the output channel signals in the value of a parameter of an acoustic signal arriving at the microphones from each of the
sound sources, said differences being attributable to the locations of the plurality of microphones;
(c) on the basis of the band-dependent inter-channel parameter value differences for each frequency band, determining which one of the respective band-divided output channel signals in each frequency band comes from which one of the sound
sources;
(d) selecting particular band-divided output channel signals determined in step (c) to have been generated from at least one of the sound sources; and
(e) combining the selected band-divided output channel signals selected for said at least one of the sound sources in the step (d) into a resulting sound source signal from said at least one of the sound sources.
2. A method according to claim 1, wherein said differences in value of a parameter include differences in at least one of time and level of each acoustic signal reaching the respective microphones.
3. A method according to claim 2 wherein in said step (a) the divided frequency bands are chosen small enough to assure that each of the band-divided output channel signals essentially and principally comprises a component of an acoustic signal
from only one of the sound sources.
4. A method according to claim 3 in which said at least one of time and level used in step (b) is time required for a component in said each frequency band of the acoustic signal to reach said microphones from each of the sound sources, and in
which the band-dependent inter-channel parameter value differences are band-dependent inter-channel time differences which represent differences between the microphones in time required for each acoustic signal in said each frequency band to reach the
respective microphones.
5. A method according to claim 4, further including a step (f) of detecting, from the output channel signals from the respective microphones, as fullband inter-channel time differences, differences between the microphones in time required for
said each acoustic signal from each of the sound sources to reach the respective microphones; and
wherein said step (c) determines, by collating the band-dependent inter-channel time differences in each frequency band with the fullband inter-channel time differences, which one of the respective band-divided output channel signals in said each
frequency band comes from which one of the sound sources.
6. A method according to claim 5 in which step (f) comprises the steps of determining cross-correlations between the output channel signals from the respective microphones, and determining the fullband inter-channel time differences as time
differences between those output channel signals which exhibit peaks in the cross-correlations.
7. A method according to claim 6, in which one of the fullband inter-channel time differences which is closest to a time corresponding to a phase difference between components in each frequency band of the band divided output channels is defined
as the band-dependent inter-channel time difference in said each frequency band.
8. A method according to claim 3 in which said at least one of time and level used in step (b) is signal level of a component in said each frequency band of the acoustic signal arriving at each of the microphones from each of the sound sources,
and in which the band-dependent inter-channel parameter value differences represent level differences between the band divided output channel signals in said each frequency band.
9. A method according to the claim 8 in which said step (c) further comprises the steps of:
(c-1) detecting level differences between the output channel signals from the respective microphones as fullband inter-channel level differences;
(c-2) comparing a sign of each of the fullband inter-channel level differences against signs of all of the band-dependent inter-channel level differences to count the number of similar signs;
(c-3) if the number of similar signs is equal to or greater than a given number, determining that all the band-divided output channel signals corresponding to the sign of said each inter-channel level differences cone from one of the sound
sources corresponding to said sign; and
(c-4) if the number of similar signs is smaller than said given number, determining which ones of the respective band-divided output channel signals in each frequency band come from which one of the sound sources.
10. A method according to claim 3, in which said step (b) detects differences both in time for the acoustic signal in each divided frequency band to reach the microphones from each of the sound sources and in level of the acoustic signal
arriving at the microphones, wherein the band-dependent inter-channel parameter value differences include band-dependent inter-channel time differences and band-dependent inter-channel level differences, said method further comprising the steps of:
(f) detecting, from the output channel signals from the respective microphones, as inter-channel time differences, differences between the microphones in time for the acoustic signal from each of the sound sources to reach the respective
microphones; and
(g) dividing the band divided output channel signals into three frequency ranges including a low, a middle and a high range on the basis of the inter-channel time differences; and
wherein the step (c) comprises the steps of:
(c-1) determining, on the basis of the band-dependent inter-channel time differences for the frequency bands in the low range, which one of the respective band-divided output channel signals in each frequency band comes from which one of the
sound sources;
(c-2) determining, on the basis of the band-dependent inter-channel level differences and the band-dependent inter-channel time differences for the frequency bands in the middle range, which one of the respective band-divided output channel
signals in each frequency band comes from which one of the sound sources; and
(c-3) determining, on the basis of the band-dependent inter-channel level differences for frequency bands in the high range, which one of the respective band divided output channel signals in each frequency band comes from which one of the sound
sources.
11. A method according to one of claims 1, 4, 8, or 10, further comprising the steps of:
(1) detecting band-dependent levels of the output channel signals which are divided into the frequency bands;
(2) comparing, for each frequency band, the band-dependent levels between the channels, and, on the basis of a result of the comparison, detecting at least one of the sound sources which is not uttering a sound; and
(3) based on detection of a non-uttering sound source, suppressing sound source signals corresponding to said non-uttering sound source.
12. A method according to claim 11, further comprising the steps of:
(4) detecting a level of full frequency band of each of the output channels signals, thus determining a fullband level for each channel; and
(5) determining whether or not each of the fullband levels of the respective channels detected in step (4) is below a reference level, and if it is found that any one of the fullband levels is above said reference level, executing steps (1), (2)
and (3).
13. A method according to claim 12 in which in the event it is determined in step (5) that the total number of frequency bands of the highest levels is equal to or less than the reference level, all of the sound source signals produced in the
combining step (e) are suppressed.
14. A method according to claim 11 in which step (2) comprises the steps of:
(2-1) comparing band-dependent levels between the channels to determine one of the channels with a highest level for each frequency band and counting a total number of frequency bands with highest levels for each channel;
(2-2) determining, for each channel, whether or not the total number of frequency bands with the highest level exceeds a first reference value;
(2-3) if it is found in step (2-2) that one of the total numbers exceeds the first reference value, estimating, from the location of the microphone for the channel having the total number exceeding the first reference value, at least one of the
sound sources uttering a sound; and
(2-4) deciding that a sound source or sources other than the estimated sound sources are sources which are not uttering a sound.
15. A method according to claim 14, further comprising the steps of:
(2-5) in the event it is determined in step (2-2) that none of the total numbers exceeds the first reference value, determining, for each channel, if the total number of frequency bands with highest levels is equal to or less than a second
reference value which is less than the first reference value; and
(2-6) if it is determined in step (2-5) that one of the total numbers of frequency bands for that channel is less than the second reference value, deciding that at least one of the sound sources corresponding to the location of the microphone for
the channel having the total number less than the second reference value is not uttering a sound.
16. A method according to claim 15, in which the number of sound sources is equal to four or greater, and in which in the event it is determined in step (2-5) that the total number of frequency bands of the highest levels for that channel is
less than the second reference value, the second reference value is incremented in a stepwise manner consistent with a requirement that the first reference value is not exceeded by the second reference value, and repeating steps (2-5) and (2-6) a number
of times equal to or less than (M-2) where M represents the number of sound sources.
17. A method according to one of claims 1, 4, 8 or 10, further comprising the steps of:
(f) detecting time-of-arrival differences of the divided output channel signals to their associated microphones for each frequency band, thus providing band-dependent time differences;
(g) comparing the band-dependent time-of-arrival differences between the channels for each frequency band, and based on the comparison result, determining at least one of the sound sources which is not uttering a sound; and
(h) in response to a determination of the non-uttering sound source, suppressing the sound source signal corresponding to the non-uttering sound source among those sound source signals which are produced in the combining step (e).
18. A method according to claim 17, further comprising the steps of:
(i) detecting a level of full frequency band of each of the output channel signals, thus providing fullband level for each channel; and
(j) determining whether or not the fullband level of each of the channels is equal to or below a reference level, and in the event any one of the fullband levels is above the reference level, executing steps (f), (g) and (h).
19. A method according to claim 18, in which step (g) comprises the steps of:
(g-1) on the basis of the comparison of the band-dependent time-of-arrival differences for each frequency band, determining, for each frequency band, one of the channels in which an acoustic signal reached earliest and counting a total number of
frequency bands with the earliest arrivals for each channel;
(g-2) determining whether or not the total number of frequency bands with earliest arrivals in each channel exceeds a first reference value;
(g-3) in the event it is determined in step (g-2) that one of the total numbers exceeds the first reference value, estimating, on the basis of the location of the microphone for the channel having the total number exceeding the first reference
value, at least one of the sound sources as uttering a sound; and
(g-4) deciding that those sound sources other than the estimated sound source are not uttering a sound.
20. A method according to claim 19, further comprising the steps of:
(g-5) in the event it is determined in step (g-2) that none of the total numbers exceeds the first reference value, determining, for each channel, whether or not the total number of frequency bands with the earliest arrivals is below a second
reference value which is less than the first reference value; and
(g-6) in the event it is determined in step (g-5) that one of the total numbers of frequency bands is below the second reference value, determining at least one of the sound sources as not uttering a sound, on the basis of the location of the
microphone for the channel having the total number of frequency bands below the second reference value.
21. The method according to claim 20, in which the number of sound sources is equal to four or greater and in which in the event it is determined in step (g-5) that the total number is below the second reference value, the second reference value
is incremented in a stepwise manner consistent with the requirement that the first reference value is not exceeded by the second reference value, and steps (g-5) and (g-6) are repeated a number of times equal to or less than (M-2) where M represents the
number of sound sources.
22. A method according to claim 18, in which in the event it is determined in step (j) that all of the fullband levels are below the reference level, all of the sound source signals which are produced in step (e) are suppressed.
23. A method according to claim 4, further comprising the steps of:
(f) detecting a sound source which is not uttering a sound on the basis of the result of comparison of the band-dependent inter-channel time differences between the channels for each frequency band; and
(g) in response to a detection of the non-uttering sound source in step (f), suppressing a sound source signal corresponding to the non-uttering sound source among the sound source signals which are produced in step (e).
24. A method according to claim 23, further comprising the steps of:
(h) detecting a level of full frequency band of each of the output channel signals to provide a fullband level for each channel; and
(i) determining, for each channel, whether or not the fullband level detected in step (h) is below a reference level value, and in the event it is determined that any one of the fullband levels is above the reference level, steps (f) and (g) are
executed.
25. A method according to claim 24, in which step (f) comprises the steps of:
(f-1) based on the comparison of the band-dependent inter-channel time differences for each band, determining, for each band, one of the channels in which an acoustic signal arrives earliest, and counting a total number of frequency bands with
the earliest arrivals for each channel;
(f-2) determining, for each channel, whether or not the total number of frequency bands with the earliest arrivals exceeds a first reference value;
(f-3) if it is determined in step (f-2) that one of the total numbers exceeds the first reference value, estimating, from the location of the microphone for the channel having the total number exceeding the first reference value, at least one of
the sound sources uttering a sound; and
(f-4) deciding that a sound source or sources other than the estimated sound source is not uttering a sound.
26. A method according to claim 25, further comprising the steps of:
(f-5) in the event it is determined in step (f-2) that none of the total numbers exceeds the first reference value, determining, for each channel, whether or not the total number of frequency bands with the earliest arrivals is below a second
reference value which is less than the first reference value; and
(f-6) in the event it is determined in step (f-5) that one of the total numbers of frequency bands is below the second reference value, determining at least one of the sound sources as not uttering a sound, on the basis of the location of the
microphone for the channel having the total number of frequency bands below the second reference value.
27. A method according to one of claims 4, 8, 10 or 2, wherein said step (d) selects the band-divided output channel signals that come from each of the sound sources, respectively, and said step (e) combines band-divided output channel signals
selected for each of the sound sources to produce sound source signals as from the sound sources, respectively, said method further comprising the steps of:
(1) determining a power spectrum for each output channel from the respective microphone;
(2) dividing the power spectrums of all the channels into frequency bands such that each frequency band contains components of at most one of the sound sources, and detecting levels of each channel in each frequency band as a band-dependent
level;
(3) comparing the band-dependent levels in each frequency band to determine a channel exhibiting the maximum level for each frequency band;
(4) determining the status of a sound source including counting, for each channel, the number of frequency bands which exhibited the maximum levels, determining, for each channel, whether or not the number of frequency bands exhibiting maximum
levels exceeds a first reference value, and determining that a sound source or sound sources other than the sound source in a zone covered by the microphone of the channel for which the number of bands exceeds the first reference value are not uttering
acoustic sounds; and
(5) suppressing a sound source signal or sound source signals corresponding to the sound source or sound sources which is determined as not uttering acoustic sounds from among the sound source signals which are produced in step (e) signals.
28. A method according to one of claims 4, 8, 10 or 2, in which in the step (b), if a frequency range of the acoustic signal from one of the sound sources is preknown to be broader than frequency ranges of the acoustic signals from the other
sound sources, the detection of the band-dependent inter-channel parameter value differences is not executed for frequency bands in those portions of the broader frequency range other than a portion where the broader frequency range overlaps the
frequency ranges of the acoustic signals from said other sound sources, and in step (c), a determination is rendered that the band-divided output channel signals in said portions of the broader frequency range come from said preknown sound source.
29. A method according to one of claims 1, 4, 8 or 10 in which at least one of the sound sources is a speaker while at least one of the other sound sources is electroacoustical transducer means which converts a received signal oncoming from a
remote end into an acoustic signal, and in which step (d) comprises the steps of: interrupting components of an acoustic signal from the electroacoustical transducer means in the band-divided channel signals, while selecting components of an acoustic
signal from the speaker, and transmitting a sound source signal which is produced in step (e) to the remote end.
30. A method according to claim 29, further comprising the steps of:
(1) dividing a received signal from the electroacoustical transducer means into a plurality of frequency bands so that each frequency band contains a component of an acoustic signal from only one of the sound sources;
(2) determining each frequency band of the band divided received signal as a transmittable band if the level of the frequency band is below a given value; and
(3) selecting those transmittable bands to be fed to step (e).
31. A method according to claim 30, in which the selection of the transmittable bands is delayed in correspondence to a propagation time of an acoustic signal between the electroacoustical transducer means and the microphone.
32. A method according to claim 29, further comprising the steps of:
(1) dividing a received signal into a plurality of frequency bands so that each frequency band contains a component of an acoustic signal from only one of the sound sources;
(2) eliminating, from the band divided components of the received signal, the frequency band selected in step (d); and
(3) combining the remaining band components of the received signal into a signal in the time domain to be fed to the electroacoustical transducer means.
33. A method according to one of claims 1, 4, 8 or 10, further comprising the steps of:
(1) dividing each of the output channel signals from the respective microphones into another plurality of frequency bands chosen small enough to assure that each of the frequency bands contains a component of an acoustic signal from only one of
the sound sources;
(2) detecting band-dependent levels of the output channel signals in each of said another plurality of frequency bands, thereby providing
band-dependent levels;
(3) comparing the band-dependent levels between the channels for each frequency band, and detecting, on the basis of a result of the comparison, at least one of the sound sources as a non-uttering sound source which is not uttering a sound; and
(4) suppressing the sound source signal which corresponds to the non-uttering sound source among the sound source signals which are produced in step (e) in response to a detection of the non-uttering sound source in step (3).
34. A method according to claim 33, further comprising the steps of:
(5) detecting a level of a full frequency band of each of the output channel signals, thereby providing a fullband level for each channel; and
(6) determining whether or not each of the fullband levels of the respective channels is equal to or below a reference level, and in the event any one of the fullband levels is above the reference level, executing steps (1), (2) and (3).
35. A method according to claim 34, in which step (3) comprises the steps of:
(3-1) determining, for each frequency band, one of the channels in which the band-dependent level is the highest, and counting the number of frequency bands with the highest levels for each channel;
(3-2) determining, for each frequency band, a total number of frequency bands with the highest level;
(3-3) determining, for each channel, if the total number of frequency bands with the highest levels exceeds a first reference value;
(3-4) estimating at least one of the sound sources as a sound uttering sound source which is at a location covered by one of the microphones for the channel having the total number exceeding the first reference value; and
(3-5) deciding a sound source or sources other than the estimated sound source as not uttering a sound.
36. A method according to claim 35, comprising further steps of:
(7) in the event it is determined in step (3-3) that the first reference value is not exceeded by any of the total numbers, determining, for each channel, if the total number of frequency bands with highest levels is equal to or less than a
second reference value which is less than the first reference value; and
(8) detecting at least one of the sound sources as a non-uttering sound source which is at a location covered by one of the microphones for the channel having the total number determined in step (7) to be below the second reference value.
37. A method according to claim 36, in which the number of sound sources is equal to four or greater, and in which in the event it is determined in step (7) that the total number of frequency bands with the highest levels is below the second
reference value, the second reference value is incremented in a stepwise manner consistent with the requirement that the first reference value be not exceeded by the second reference value, and steps (7) and (8) are repeated a number of times equal to or
less than (M-2) where M represents the number of sound sources.
38. A method according to claim 34 in which in the event it is determined in step (6) that the total number of frequency bands is equal to or less than the reference level, all of the sound source signals which are produced in step (e) are
suppressed.
39. A method according to one of claims 1, 4, 8 or 10, further comprising the steps of:
(1) dividing each of the output channel signals from the microphones into band-divided output channel signals of a second plurality of frequency bands chosen small enough to assure that each second band-divided output channel signal contains
essentially and principally a component of an acoustic signal from only one of the sound sources;
(2) detecting time-of-arrival differences of the respective second band-divided output channel signals to their associated microphones for each frequency band, thus providing band-dependent time differences;
(3) comparing the band-dependent time-of-arrival differences between the channels for each frequency band, and, based on the comparison result, detecting at least one of the sound sources as a non-uttering sound source which is not uttering a
sound; and
(4) in response to a detection of the non-uttering sound source by step (3), suppressing the sound source signal corresponding to the non-uttering sound source among the sound source signals which are produced in step (e).
40. A method according to claim 39, further comprising the steps of:
(5) detecting a level of full frequency band of each of the respective output channel signals, thus providing a fullband level for each channel; and
(6) determining whether or not each of the fullband levels of the respective channels is equal to or below the reference level, and transferring to step (3) if any one of the fullband levels is not below the reference level.
41. A method according to claim 40, in which step (3) comprises the steps of:
(3-1) on the basis of the comparison of the band-dependent time-of-arrival differences for each frequency band, determining, for each frequency band, one of the channels in which an acoustic signal is reached earliest;
(3-2) determining, for each channel, if the total number of frequency bands with earliest arrivals in each channel exceeds a first reference value;
(3-3) assuming at least one of the sound sources as an uttering sound source that is at a location covered by one of the microphones for the channel having the total number exceeding the first reference value; and
(3-4) determining a sound source or sources other than the assumed sound sources as not uttering a sound.
42. A method according to claim 41, further comprising the steps of:
(3-5) in the event it is determined in step (3-2) that there is no total number that exceeded the first reference value, determining whether or not the total numbers of frequency bands with earliest arrivals are below a second reference value
which is smaller than the first reference value; and
(3-6) detecting any one of the sound sources as a non-uttering sound source which is at a location covered by one of the microphones for the channel having the total number determined in step (3-5) to be below the second reference value.
43. A method according to claim 42, in which the number of sound sources is equal to four or greater, and in which in the event it is determined in step (3-5) that the total number of frequency bands with earliest arrivals is below the second
reference value, the second reference value is incremented in stepwise fashion consistent with the requirement that the first reference value be not exceeded by the second reference value, and steps (3-5) and (3-6) are repeated a number of times equal to
or less than (M-2) where M represents the number of sound sources.
44. A method according to claim 40, in which if it is determined in step (6) that all of the fullband levels are equal to or less than the reference level, all of the sound source signals which are produced in step (e) are suppressed.
45. A method according to claim 11, in which at least one of the sound sources is a speaker while at least one of the other sound sources is electroacoustical transducer means which converts a signal oncoming from a remote end into an acoustic
signal, and step (d) comprises a step of interrupting components of an acoustic signal from the electroacoustical transducer means in the band-divided channel signals while selecting components of an acoustic signal from the speaker, and transmitting a
sound source signal which is produced in step (e) to the remote end.
46. A method according to claim 45, further comprising the steps of:
(4) dividing the received signal from the electroacoustical transducer means into a plurality of frequency bands such that each frequency band contains a component of an acoustic signal from only one of the sound sources;
(5) determining each frequency band of the band-divided received signal as a transmittable band if the level of the frequency band is equal to or less than a given value; and
(6) selecting only those transmittable bands to be fed to the sound source combining step (e).
47. A method according to claim 46, further comprising a step of delaying the selection of the transmittable bands in correspondence with a propagation time of an acoustic signal between the electroacoustical transducer means and the microphone.
48. A method according to claim 45, comprising further steps of:
(4) dividing the received signal into a plurality of frequency bands so that each frequency band contains a component of an acoustic signal from only one of the sound sources;
(5) eliminating the bands selected in step (d) from the band divided components of the received signal; and
(6) combining the remaining band components in the received signal into a signal in the time domain to be supplied to the electroacoustical transducer means.
49. A method of separating at least one sound source from a plurality of sound sources by using a plurality of microphones located in spaced relation to each other, comprising the steps of:
(a) determining power spectrums for output channel signals from the respective microphones;
(b) dividing the power spectrum of each channel into a plurality of frequency bands so that principally spectrum components from a single one of the sound sources are contained in each band;
(c) detecting, for each band, differences in the divided power spectrums between the channels as band-dependent inter-channel level differences;
(d) on the basis of the band-dependent inter-channel level differences for the respective bands, determining which one of the respective divided power spectrums in each frequency band comes from which one of the sound source signals;
(e) on the basis of a determination rendered in step (d), selecting particular band divided spectrums of at least one of the channels corresponding to at least one of the sound sources; and
(f) combining the selected band divided spectrums selected in step (e) into a resulting sound source signal.
50. A method according to claim 49, further comprising the steps of:
(g) detecting level differences between the output channel signals from the respective microphones as fullband inter-channel level differences;
(h) comparing a sign of each of the fullband inter-channel level differences against signs of all of the band-dependent inter-channel level differences to count the number of similar signs;
(i) if the number of similar signs is equal to or greater than a given number, determining that all the band divided output channel signals corresponding to the sign of said each inter-channel level difference come from one of the sound sources
corresponding to said sign; and
(j) if the number of similar signs is smaller than said given number, determining which ones of the respective band-divided output channel signals in each frequency band come from which one of the sound sources.
51. An apparatus for separating at least one sound source from a plurality of sound sources using a plurality of microphones disposed in spaced relation to one another comprising:
band dividing means for dividing an output channel signal from each of the respective microphones into a plurality of frequency bands to produce band-divided output channel signals such that each of the band-divided output channel signals
essentially and principally comprises a component of an acoustic signal from only one of the sound sources;
means for detecting, for each frequency band, as band-dependent inter-channel parameter value differences, differences between the output channel signals in the value of a parameter of an acoustic signal arriving at the microphones from each of
the sound sources, said differences being attributable to the locations of the plurality of microphones;
means for determining, on the basis of the band-dependent inter-channel parameter value differences for each frequency band, which one of the respective band-divided output channel signals in each frequency band comes from which one of the sound
sources;
selecting means for selecting particular band-divided output channel signals determined by the determining means to have been generated from at least one of the sound sources; and
combining means for combining the selected band-divided output channel signals selected by said selecting means into a resulting sound source signal from said at least one of the sound sources.
52. An apparatus according to claim 51, wherein said differences in value of a parameter include differences in at least one of time and level of each acoustic signal reaching the respective microphones.
53. An apparatus according to claim 52, in which said at least one of time and level used for detecting the band-dependent inter-channel parameter value differences is a time required for a component in said each frequency band of the acoustic
signal to reach each microphone from each of the sound sources, and the band-dependent inter-channel parameter value differences are band-dependent inter-channel time differences between the microphones required for each acoustic signal in said each
frequency band to reach the respective microphones.
54. An apparatus according to claim 52, further comprising
means for detecting, from the output channel signals from the respective microphones, as fullband inter-channel time differences between the microphones, the time required for each acoustic signal from each of the sound sources to reach the
respective microphones; and
said means for determining a sound source signal comprises means for collating the band-dependent inter-channel time differences in each frequency band with the fullband inter-channel time differences to determine which one of the respective
band-divided output channel signals in said each frequency band comes from which one of the sound sources.
55. An apparatus according to claim 52, in which said at least one of time and level used by said means for detecting the band-dependent inter-channel parameter value differences is signal level of a component in said each frequency band of the
acoustic signal arriving at each of the microphones from each of the sound sources, and the band-dependent inter-channel parameter value differences are band-dependent inter-channel level differences between the band-divided output channel signals in
said each frequency band.
56. An apparatus according to claim 55, further comprising:
means for detecting level differences between the output channel signals from the respective microphones as fullband inter-channel level differences;
means for comparing a sign of each of the fullband inter-channel level differences against signs of all of the band-dependent inter-channel level differences to count the number of similar signs; and
means for determining, if the number of similar signs is equal to or greater than a given number, that all the band-divided output channel signals corresponding to the sign of said each inter-channel level differences are from one of the sound
sources corresponding to said sign, and for determining, if the number of similar signs is smaller than said given number, which ones of the respective band-divided output channel signals in each frequency hand come from which one of the sound sources.
57. An apparatus according to claim 52, in which said means for detecting band-dependent inter-channel parameter value differences detects differences both in time required for the acoustic signal in each frequency band to reach the microphones
from each of the sound sources and in level of the acoustic signal arriving at the microphones, and the band-dependent inter-channel parameter value differences including band-dependent inter-channel time differences and band-dependent inter-channel
level differences, said apparatus further comprising:
means for detecting, from the output channel signals from the respective microphones as inter-channel time differences, differences in time for the acoustic signal from each of the sound sources to reach the respective microphones; and
range dividing means for dividing the band-divided output channel signals into three frequency ranges including a low, a middle, and a high range on the basis of the inter-channel time differences, and
wherein said means for determining the sound source signal comprises:
means for determining, on the basis of the band-dependent inter-channel time differences for the frequency bands in the low range, which one of the respective band-divided output channel signals in each frequency band comes from which one of the
sound sources;
means for determining, on the basis of the band-dependent inter-channel level differences and band-dependent inter-channel time differences for the frequency bands in the middle range, which one of the respective hand-divided output channel
signals in each frequency band comes from which one of the sound sources; and
means for determining, on the basis of the band-dependent inter-channel level differences for frequency bands in the high range, which one of the respective band-divided output channel signals in each frequency band comes from which one of the
sound sources.
58. An apparatus according to one of claims 51, 53, 55 or 57, further comprising:
means for detecting the band-dependent levels of the output channel signals which are divided into frequency bands;
means for determining the status of a sound source by comparing, for each frequency band, the band-dependent levels between the channels, and detecting, on the basis of comparison result, at least one of the sound sources as a non-uttering sound
source which is not uttering a sound; and
means for suppressing, in response to the detection of the non-uttering sound source, one of the sound source signals corresponding to said at least one of the sound sources.
59. An apparatus according to claim 58, further comprising:
a fullband level detecting means for detecting a level of full frequency band of each output channel signal as a fullband level for each channel;
decision means for determining whether or not each of the fullband levels of the respective channels detected by the fullband level detecting means is below a reference level, and if any one of the fullband levels is determined to be above the
reference level, effecting the operations of said means for detecting the band-dependent levels, said means for determining the status of the sound source, and said means for suppressing.
60. An apparatus according to claim 58, in which said means for determining the status of a sound source comprises:
means for comparing the band-dependent level difference between the channels to determine one of the channels with the highest level for each frequency band, and counting the number of frequency bands with highest levels for each channel;
means for determining a total number of frequency bands with the highest levels;
decision means for determining, for each channel, whether or not the total number of frequency bands with the highest | | |
