as) United States
`a2) Patent Application Publication 10) Pub. No.: US 2011/0103626 Al
`Bisgaardet al.
`(43) Pub. Date:
`May5, 2011
`
`
`US 20110103626A1
`
`(30)
`
`Foreign Application Priority Data
`
`Publication Classification
`
`(73) Assignee:
`
`(51)
`
`Int.Cl.
`HOAR 25/00
`
`2006.01
`
`)
`(
`(52) US. CD. eee ceceeectesesenscneecessescneneeansenees 381/313
`
`(54) HEARING INSTRUMENT WITH ADAPTIVE
`DIRECTIONAL SIGNAL PROCESSING
`Jun. 23, 2006=(DK) wee PA 2006 00852
`(75)
`Inventors:
`Nikolai Bisgaard, Lyngby (DK);
`Rob Anton Jurjen DE Vries,
`Tilburg (NL)
`GN RESOUNDASS, Ballerup
`(DK)
`12/306,515
`
`(22)
`
`Jun. 25, 2007
`
`(21) Appl. No.:
`.
`PCTFiled:
`;
`PCT/DK07/00308
`(86) PCT No.:
`1
`371
`Mar.
`27, 2009
`‘ nen »
`area's
`(2), (4)
`Date:
`+
`as
`Related U.S. Application Data
`(60) Provisional application No. 60/816,244, filed on Jun.
`23, 2006.
`
`ABSTRACT
`(57)
`A hearing instrumentincludes a signal processor, and at least
`two microphonesfor reception ofsound and conversion ofthe
`received sound into corresponding electrical sound signals
`that are input to the signal processor, wherein the signal
`processoris configured to process the electrical sound signals
`into a combinedsignal with a directivity pattern with at least
`one adaptive null direction @, and wherein the signal proces-
`sor is further configured to preventthe at least one null direc-
`tion 6 from entering a prohibited range ofdirections, wherein
`the prohibited range is a function of a parameter of the elec-
`trical sound signals.
`
`an eee eee eee 7
`
`
`
`|
`
`DATA
`
`RAM
`
`|
`:
`:
`|
`
`|
`
`DATA
`
`ROM
`
`5
`
`12
`
`4g
`
`7
`|
`!
`1
`aan
`;
`2
`!
`|
`|
`/
`7
`ae
`i
`:
`: C)
`AD
`|
`|
`\
`|
`a a |
`!
`|
`D/A
`5
`,
`j
`|
`po
`| Oa
`Po
`oweb fe !
`|
`2b
`
`i
`
`i
`
`——
`=
`
`|
`
`||
`
`| |
`
`|
`
`.
`
`‘
`
`EEPROM
`
`APPLE 1010
`
` 1
`
`APPLE 1010
`
`1
`
`

`

`Patent Application Publication
`
`May5, 2011 Sheet 1 of 2
`
`US 2011/0103626 Al
`
` odFOVAYSLNI
`
`:WowWH||wwepoud
`WVuDOUd
`
`2
`
`

`

`Patent Application Publication
`
`US 2011/0103626 Al
`
`May 5, 2011 Sheet 2 of 2
`
`
`
`
`
`
`
`
`3
`
`

`

`US 2011/0103626 Al
`
`May5, 2011
`
`HEARING INSTRUMENT WITH ADAPTIVE
`DIRECTIONAL SIGNAL PROCESSING
`
`RELATED APPLICATION DATA
`
`[0001] This application is the national stage of Interna-
`tional Application No. PCT/DK2007/000308, filed on Jun.
`25, 2007, which claimspriority to andthe benefit of Denmark
`Patent Application No. PA 2006 00852, filed on Jun. 23, 2006,
`and U.S. Provisional Patent Application No. 60/81 6,244, filed
`on Jun. 23, 2006, the entire disclosure of all of which is
`expressly incorporated by reference herein.
`
`FIELD
`
`[0002] The present application relates to a hearing instru-
`ment, such as a hearing aid, an implantable hearing prosthe-
`sis, ahead set, a mobile phone,etc, with a signal processorfor
`directional signal processing.
`
`BACKGROUND
`
`Itis well-knownto use informationonthe directions
`[0003]
`to sound sources in relation to a listener for distinguishing
`between noise sources and desired sound sources. Through-
`out the present specification, the term directional signal pro-
`cessing system means a signal processing system that is
`adapted to exploit the spatial properties of an acoustic envi-
`ronment. Directional microphonesare available, but typically
`directional signal processing systems utilize an array of
`omni-directional microphones.
`[0004] The directional signal processing system combines
`the electrical signals from the microphonesin the array into a
`signal with varying sensitivity to sound sources in different
`directions in relation to the array. Throughout the present
`specification, a plot of the varying sensitivity as a function of
`the direction is denoted the directivity pattern. Typically, a
`directivity pattern has at least one direction wherein the
`microphonesignals substantially cancel each other. Through-
`out the present specification, such a direction is denoted a null
`direction. A directivity pattern may comprise several null
`directions depending on the number of microphones in the
`array and dependingonthe signal processing.
`[0005] Directional signal processing systems are known
`that prevent sound suppression of sources in certain direc-
`tions of interest.
`[0006]
`For example, U.S. Pat. No. 5,473,701 discloses a
`method of enhancing the signal-to-noise ratio of a micro-
`phone array with an adjustable directivity pattern, i.e. an
`adjustable null direction, for reduction of the microphone
`array output signal
`level
`in accordance with a criterion
`wherein the reduction is performed undera constraint that the
`null direction is precluded from being located within a pre-
`determined region of space.
`[0007]
`Itis an object to provide a system with an improved
`capability of suppressing sound sources from all directions.
`
`SUMMARY
`
`[0008] According to the present application, the above-
`mentioned and other objects are fulfilled by a hearing instru-
`ment with at least two microphones for reception of sound
`and conversion ofthe received sound into correspondingelec-
`trical sound signals that are input to a signal processor,
`wherein the signal processor is adapted to process the elec-
`trical sound signals into a combinedsignal with a directivity
`pattern with at least one adaptive null direction 0. The signal
`
`processoris further adapted to prevent the at least one adap-
`tive null direction @ from entering one or more prohibited
`rangesofdirections, wherein each prohibited range is a func-
`tion of a parameterof the electrical soundsignals.
`[0009] More than one prohibited range may for example
`occur in situations with more than onedesired signal arriving
`from different directions.
`
`Preferably, the at least two microphones are omni-
`[0010]
`directional microphones; however in some embodiments,
`some of the at least two microphonesare substituted with
`directional microphones.
`[0011]
`It is an important advantage that suppression of
`desired sound sources are avoided while undesired sound
`
`sources maystill be suppressed from anyarbitrary direction.
`[0012] The hearing instrument may further comprise a
`desired signal detector for detection of desired signals, for
`example a speech detector for detection of presence of
`speech. Adjustmentofthe prohibited range of directions may
`be performed gradually overa first time interval when desired
`signals, such as speech,are detected after a period of absence
`of speech.
`[0013]
`Further, adjustment of the prohibited range(s) of
`directions may be performed gradually over a second time
`interval when a desired signal, such as speech, stops after a
`period of presence of the desired signal, e.g. speech.
`[0014] The prohibited range may include a predetermined
`direction, such as 0° azimuth or another preferred direction.
`[0015] An estimate of the power of sound received by at
`least one of the at least two microphones may constitute the
`parameter, for example the averaged powerof sound received
`by a front microphone mayconstitute the parameter, or the
`parameter may bea function of the estimate of the powerof
`sound, e.g. the averaged power of sound.
`[0016] An estimate of the signal to noise ratio of sound
`received by at least one ofthe at least two microphones may
`constitute the parameter, or the parameter may be a function
`of the estimate of the signal to noiseratio.
`[0017] The hearing instrument may further comprise a
`desired signal detector, such as a speech detector, and a direc-
`tion of arrival detector, and the prohibited range may include
`the detected direction of arrival of a detected desired signal,
`such as speech, whereby suppression ofthe desiredsignal, is
`prevented.
`[0018]
`In an embodiment with a single prohibited range,
`the prohibited range may, in the presence of multiple desired
`signal sources, such as multiple speech sources, include the
`detected direction of arrival of the detected desired signal
`source closest to 0° azimuth, or another preferred direction.
`[0019]
`In an embodimentwith a single prohibited range,
`the prohibited range may, in the presence of multiple desired
`signal sources, such as speech sources, include the detected
`directionsofarrival ofall desired signal sources.
`[0020]
`In an embodiment with a plurality of prohibited
`ranges, someorall of the prohibited ranges may be centered
`on respective detected directions of desired signal sources.
`[0021] As explained fora single prohibited range, the width
`of a specific prohibited range of the plurality of prohibited
`ranges centered on the corresponding direction of the corre-
`sponding desired signal source may be controlled as a func-
`tion of a parameterofthe electrical sound signals, e.g. power,
`signal-noise ratio, etc.
`[0022] A current null direction mayreside inside the pro-
`hibited range(s) of directions upon adjustmentofthe prohib-
`
`4
`
`

`

`US 2011/0103626 Al
`
`May5, 2011
`
`ited range(s) of directions. The signal processor may further
`be adapted to move such a null direction outside the adjusted
`prohibited range(s).
`[0023] The signal processor may be configured for subband
`processing whereby the electrical sound signals from the
`microphonesare dividedintoa set of frequency bandsB, and,
`in each frequency band B, or at least in someofthe frequency
`bands B, the electrical sound signals are individually pro-
`cessed including:
`[0024]
`(1) processing the electrical signals into a combined
`signal with an individual directivity pattern with an individu-
`ally adapted null direction 0,, and
`[0025]
`(2) preventing the null direction 0, from entering one
`or more prohibited rangesof directions, wherein each prohib-
`ited range is a function of a parameterofthe electrical sound
`signals.
`Subband processing allows individual suppression
`[0026]
`of undesired sound sources emitting sound in different fre-
`quency ranges.
`[0027] The signal processor may be adapted to perform
`directional signal processing selected from the group consist-
`ing of an adaptive beam former, a multi-channel Wienerfilter,
`an independent componentanalysis, and a blind source sepa-
`ration algorithm.
`[0028]
`In accordance with some embodiments, a hearing
`instrumentincludes a signal processor, and at least two micro-
`phonesfor reception of sound and conversionofthe received
`sound into corresponding electrical sound signals that are
`input to the signal processor, wherein the signal processoris
`configured to process the electrical sound signals into a com-
`bined signal with a directivity pattern with at least one adap-
`tive null direction 6, and wherein the signal processor is
`further configured to preventthe at least one null direction 8
`from entering a prohibited range of directions, wherein the
`prohibited rangeis a function of a parameteroftheelectrical
`sound signals.
`
`DESCRIPTION OF THE DRAWING FIGURES
`
`[0029] The above and other features and advantages will
`become more apparent to those of ordinary skill in the art by
`describing in detail exemplary embodiments thereof with
`reference to the attached drawings in which:
`[0030]
`FIG. 1 showsa simplified block diagram ofa digital
`hearing aid according to some embodiments, and
`[0031]
`FIG. 2 schematically illustrates the directional sig-
`nal processing of the hearing aid of FIG.1.
`
`DETAILED DESCRIPTION
`
`[0032] The embodiments will now be described more fully
`hereinafter with reference to the accompanying drawings.
`The claimed invention may, however, be embodiedin differ-
`ent forms and should not be construed as limited to the
`embodiments set forth herein. Thus, the illustrated embodi-
`ments are not intended as an exhaustive description of the
`invention or as a limitation on the scope of the invention. In
`addition, an illustrated embodiment needs not have all the
`aspects or advantages shown. An aspect or an advantage
`described in conjunction with a particular embodimentis not
`necessarily limited to that embodiment and can bepracticed
`in any other embodiments even if not so illustrated. Like
`reference numerals refer to like elements throughout.
`[0033]
`FIG. 1 showsa simplified block diagram ofa digital
`hearing aid according to some embodiments. The hearing aid
`
`1 comprises one or more soundreceivers 2, e.g. two micro-
`phones 2a anda telecoil 26. The analog signals for the micro-
`phones are coupled to an analog-digital converter circuit 3,
`which contains an analog-digital converter 4 for each of the
`microphones.
`[0034] The digital signal outputs from the analog-digital
`converters 4 are coupled to a commondata line 5, which leads
`the signals to a digital signal processor (DSP) 6. The DSPis
`programmed to perform the necessary signal processing
`operations of digital signals to compensate hearing loss in
`accordance with the needs of the user. The DSP is further
`
`programmed for automatic adjustment of signal processing
`parameters in accordance with some embodiments.
`[0035] The output signal is then fed to a digital-analog
`converter 12, from which analog output signals are fed to a
`sound transducer 13, such as a miniature loudspeaker.
`[0036]
`In addition, externally in relation to the DSP 6, the
`hearing aid contains a storage unit 14, which in the example
`shown is an EEPROM(electronically erasable programmable
`read-only memory). This external memory 14, which is con-
`nected to a commonserial data bus, can be provided via an
`interface 15 with programmes, data, parameters etc. entered
`from a PC 16, for example, when a new hearingaidis allotted
`to a specific user, where the hearing aid is adjusted for pre-
`cisely this user, or when a user has his hearing aid updated
`and/or re-adjusted to the user’s actual hearingloss, e.g. by an
`audiologist.
`[0037] The DSP 6 contains a central processor (CPU)7 and
`a numberof internal storage units 8-11, these storage units
`containing data and programmes, which are presently being
`executed in the DSP circuit 6. The DSP 6 contains a pro-
`gramme-ROM(read-only memory) 8, a data-ROM 9, a pro-
`gramme-RAM (random access memory) 10 and a data-RAM
`11. The twofirst-mentioned contain programmes and data
`which constitute permanent elements in the circuit, while the
`twolast-mentioned contain programmesand data which can
`be changed or overwritten.
`[0038] Typically, the external EEPROM 14 is considerably
`larger, e.g. 4-8 times larger, than the internal RAM, which
`meansthat certain data and programmescan bestored in the
`EEPROMsothat they can be read into the internal RAMsfor
`execution as required. Later, these special data and pro-
`grammes maybe overwritten by the normal operational data
`and working programmes. The external EEPROM can thus
`contain a series of programmes, which in some embodiments
`are used only in special cases, such as e.g. start-up pro-
`grammes.
`[0039]
`FIG. 2 schematically illustrates the signal process-
`ing of a hearing instrument according to some embodiments.
`Theillustrated hearing instrument has two microphones20,
`22 positioned in a housing to be worn at the ear of the user.
`Whenthe housing is mounted in its operating position at the
`ear of the user, one of the microphones, the front microphone
`20, is positioned in front of the other microphone, the rear
`microphone 22, and a horizontal line extending through the
`front and rear microphones defines the front direction,i.e.
`azimuth=0°, corresponding to the looking direction of the
`user of the hearing instrument.
`[0040]
`In another embodiment comprising a binaural hear-
`ing aid, the microphones20, 22 may be positionedin separate
`housings, namely a housing positioned in the left ear and a
`housing positioned in the right ear ofthe user. The directional
`signal processing may then take place in either of the left or
`right hearing aid housings, or in both housing, or in a separate
`
`5
`
`

`

`US 2011/0103626 Al
`
`May5, 2011
`
`housing containing signal processing circuitry and intended
`to be worn elsewhere on the body of the user. Theelectrical
`signals may be communicated between the housings with
`electrical wires or wirelessly. The large distance between
`microphonesin the left ear housing and the right ear housing
`maylead to a directivity pattern with a large directivity.
`[0041] The microphones 20, 22 convert received sound
`signals into corresponding electrical sound signals that are
`converted into digital soundsignals 24, 26 by respective A/D
`converters (not shown).
`[0042] Each ofthe digitized soundsignals 24, 26 is input to
`a respective subtraction circuit 28, and a respective delay 32,
`34 with delay D,,. Each delay 32, 34 delays the digitized
`sound signal 24, 26 by the amount of time used by a sound
`signal to propagate in the 0° azimuth direction from the front
`microphone 20 to the rear microphone 22. Each subtraction
`circuit 28, 30 subtracts the respective delayed signal 36, 38
`from one microphone 20, 22 from the direct signal 26, 24 of
`the other microphone 22, 20. Each of the subtracted signals
`40, 42 has a fixed directional pattern 44, 46, a so-called
`cardioid pattern. The cardioid pattern 44 of the upper branch
`(a) has a null direction 48 at 180° azimuth, i.e. pointing in the
`rear direction of the user, and the cardioid pattern 46 of the
`lower branch (b) has a null direction 50 at 0° azimuth, i.e.
`pointing in the front direction ofthe user.
`[0043] The subtracted signal 42 of the lower branch (b)is
`filtered by an adaptive filter 52 with a transfer function H, and
`the subtracted signal 40 of the upper branch(a) is delayed by
`a delay 54 with a delay D,, equal to the group delay of the
`adaptivefilter 52, and subsequently the two signals 56, 58 are
`subtracted for formation of a combined signal 60 with a
`directivity pattern 62 with an adaptive null direction 0. An
`example of a resulting directivity pattern 62 is also shown in
`FIG.2. The hatchedarea ofthe resulting directivity pattern 62
`illustrates the prohibited range of directions which in the
`illustrated example is symmetrical around 0° azimuth. The
`arched arrowsindicate that the prohibited range ofdirections
`vary as a function of a parameter of the electrical sound
`signals.
`It should be notedthat in the illustrated embodiment
`[0044]
`of FIG. 2, the delay 34 and the subtraction circuit 28 may be
`omitted and still an output 60 with a directional pattern 62
`similar to the illustrated embodiment of FIG. 2 may be
`obtained due to corresponding changesin the operation ofthe
`adaptivefilter 52.
`[0045]
`Further, both delays 32, 34 and subtraction circuits
`28, 30 may be omitted in the illustrated embodimentof FIG.
`2, andstill an output 60 with a directional pattern 62 similar to
`the illustrated embodiment of FIG. 2 may be obtained due to
`corresponding changes in the operation of the adaptivefilter
`52.
`
`gain G,,, and the adaptive filter controller 66 constrains the
`gain G,, to remain inside the range O=G,=G,,,,,;,- The value
`of the threshold G,,,,;, determines the prohibited range of
`directions. For example, when G,,,,,=1, the prohibited range
`of directions ranges from -90° azimuth to +90° azimuth.
`[0050] The adaptivefilter controller 66 mayfreezethefilter
`coefficients, i.e. updating of the filter coefficients may be
`stopped temporarily, when the strongest sound source is
`located within the prohibited range of directions. This
`approach requires estimation of the direction of arrival
`(DOA)ofthe signal incident on the hearing instrument.
`[0051] A DOA estimate maybe obtained by determination
`of an M point auto-correlation A of the front microphone
`signal 24 delayed by D and determination of an M point
`cross-correlation B ofthe front microphonesignal 24 delayed
`by D andthe rear microphonesignal 26:
`
`M-1
`A= » (front(k — D — iP
`i=0
`
`M-1
`B= » front(k — D — i)rear(k — i)
`i=0
`
`(1)
`
`(2)
`
`B=B/A can be used asan estimateofthe direction ofarrival of
`the dominant soundin the acoustic environment. When B=B/
`A=1, the DOAis 0°. As B decreases toward 0, the DOA moves
`towards 180° azimuth. Thus, the adaptation may be tempo-
`rarily stopped when
`p>o
`
`(3)
`
`where o is determined in such a way that B=B/A=o when the
`DOA of e.g. a zero mean white noise source is a degrees
`azimuth, the prohibited range of directions extending from
`-a degrees azimuth to a degrees azimuth including 0° azi-
`muth.
`
`It should be noted that with this DOA estimate, the
`[0052]
`prohibited range of directions will be frequency dependent,
`because the value of B=B/A is both dependenton the direction
`of arrival and on the frequency ofthe signal. In some embodi-
`ments, with subband processing with individual beamform-
`ing in each frequency bandB,, individual thresholds a may be
`defined for each frequency bandB;.
`[0053] The person skilled in the art will recognize that
`numerous other conventional methodsare available to obtain
`
`an estimate of the DOA, including frequency independent
`estimates.
`
`[0054] The signal processing is not necessarily done on the
`same apparatus that contains (one or more of) the micro-
`phones. The signal processing may be performed ina separate
`In the illustrated embodiment, the filter 52 is con-
`[0046]
`devicethatis linked to the, possibly multiple apparatuses that
`figured to minimize the output powerof the combined signal
`contain the microphonesvia a wire, wireless or other connec-
`60 bythe filter coefficient update circuit 64. Thefilter 52 may
`tion.
`beafinite impulse response (FIR) filter with N taps.
`[0047] The adaptive filter controller 66 prevents the null
`[0055]
`In the following various examples are described of
`direction @ from entering a prohibited rangeof directions as a
`determining the prohibited range of directions as a function of
`function of a parameterof the electrical sound signals.
`a parameterof the electrical sound signals. In the examples,
`-c till a degrees azimuth constitutes the prohibited range of
`[0048] The adaptivefilter controller 66 constrains the filter
`directions including 0° azimuth.
`coefficients of the adaptive filter 52 in such a way that a
`directional null 6 remains outside the prohibited range of
`the prohibited range of
`[0056]
`In some embodiments,
`directions.
`directions is a function of the short term average power P,.
`(e.g. overthe past 10 seconds)ofthe electrical signal 24 from
`front microphone 20 in accordance with
`
`For example, the adaptivefilter 52 may have a single
`[0049]
`tap in which case the adaptive filter 52 is an amplifier with a
`
`6
`
`

`

`US 2011/0103626 Al
`
`May5, 2011
`
`max(Pmins min(Pr, Pmax)) — Pmin
`)
`(
`max(Prmin, min(Pr, Pmax)) — nt)
`A)
`@ = Onax —
`(Qmax — Esnr)
`@ = Qqx{ 1 ——_--AA
`Prax — Prin
`Prax — Prin
`
`[0057] Hence, the prohibited range of directions narrows
`max
`a=0°
`when the signal power P,. increases and for P,>P.
`(front direction) and for P,<P,,,,,, 2A=Qnax-
`[0058] Thevalues ofa... Pin, andP,,,.,may be set during
`manufacture of the hearing instrument, or, during a fitting
`session of the hearing instrument with the intended user.
`[0059]
`In one example, P,,,,,=45 dB<p, and P,,,,,=110
`dB.p,. It should be noted that very loud sounds (above 110
`dB.p;) may be suppressed from any direction providing pro-
`tection against harmful sounds(e.g. when getting too close to
`a loudspeaker at a concert). For a.,,,.=180°, an omni-direc-
`tional pattern is obtainedin relative quiet environments below
`45 dBxpr.
`[0060]
`Inother embodiments, the prohibited rangeofdirec-
`tions is a function of the signal-to-noise ratio SNR for the
`signal 40 at point (a) in FIG. 2 in accordance with
`
`_-
`Oe Emax
`
`max(SNRynins min(SNR, SNRinax)) — SNRonin
`SNRnax — SNR
`
`(5)
`
`[0061] Hence, the prohibited range of directions narrows
`when the signal-to-noise ration SNR increases and for
`SNROSNR,,,ox O=Qjng,, and for SNR<SNR,,,,,,. a=0° (front
`direction).
`[0062] Thevalues ofa,,,. SNR,,,,,., and SNR,,,, may be set
`during manufacture of the hearing instrument, or, during a
`fitting session of the hearing instrument to the intended user.
`[0063]
`SNR may be estimated utilizing a speech detector
`68, e.g. a modulation or speech probability estimator, or a
`modulation or speech activity detector, to detect presence of
`speech and calculate the average power P, ofthe signal when
`speech is present. The average noise powerP,, in absence of
`speechis estimated using a minimumstatistics approach. An
`estimate of the SNRis then given by
`
`SNR= 20log,o|
`
`
`Py — Py
`
`) aB
`
`(6)
`
`the prohibited range of
`In some embodiments,
`[0064]
`directions is a function of the azimuth direction of speech p.
`Presence of speech is detected by the speech detector 68 that
`processes the signal 24 and the direction ofarrival f is esti-
`mated by the direction of arrival detector 70, and the prohib-
`ited rangeofdirectionsis adjusted to include B. 6 may change
`due to head or speaker movement. In the presence of multiple
`speech sources, the prohibited range of directions may be
`adjusted to include DOAof the speech source closest to 0°
`azimuth or to include DOAsofall detected speech sources.
`[0065] The above-mentioned approaches may be com-
`bined.
`
`For example, in some embodiments, the prohibited
`[0066]
`range of directions is a function of the short term average
`powerP,. (e.g. over the past 10 seconds) of the electrical
`signal 24 from front microphone 20 in accordance with
`
`whichis similar to equation (4) above with the exception that
`TAX
`‘SMP
`a varies between G.,,,,
`and a,,,,.
`in equation (7) while a varies
`between «,,,,,, and 0° in equation (4), and wherein
`
`max(SNRmmins min(SNR, SNRmax)) — SNRinin
`SNRnax — SNRnin
`
`max(SNRiowiid, min(SNR, SNRpw) — SNRiow
`SNRiowthia — SNRiow
`
`(180 — Qinin) +
`
`(180 - ainin)
`
`(8)
`
`snr =
`
`Qmin +
`
`and
`
`(9)
`
`Onin=DOAmace
`wherein
`SNRis the estimated signal-to-noise ratio at point (a) in FIG.
`1 over the past 10 seconds, e.g. obtained as described above,
`SNR,,,o72 18 the estimated signal-to-noise ratio at point (a) in
`FIG.1 overthe past 0.05 seconds, e.g. obtained as described
`above,
`is the maximum value of SNR.short over the past
`SNRsjoramax
`10 seconds, and
`DOA,,,.. 18 the average value of the DOA overthe 0.05 sec-
`onds block that resulted in SNRyyoremax
`[0067]
`For example, P,,,,.=60 dBspr, Prmin=45 dBsp;;
`SNR,,,,=) dB, SNR,,,,=15 dB, SNR,,,,=-10 dB, SNR-
`lowthid=—20 GB, and o.,,,,=180°.
`[0068]
`It should be noted that in this embodimentthe pro-
`hibited rangeofdirections is as narrow as possible around the
`direction to the speech source with the highest SNR. The
`prohibited range increases when the overall SNR is larger
`than the threshold SNR,,,,,, or smaller than the threshold SNR-
`iow: and saturates into an omni-directional pattern when the
`SNRis larger than the threshold SNR,,,,.,. (e.g. when there is
`no noise) or lower than the threshold SNR,,,,,,:¢ (e-g- when
`there is no speech), or the overall signal power P;. is smaller
`than the threshold P,,,,, and also saturates into an omni-
`directional pattern when P,. is smaller than the threshold P,,,,,,
`(e.g. in quiet surroundings).
`[0069]
`Preferably, timing restrictions are also included in
`accordance with some embodiments so that frequent and
`abrupt changes of the prohibited range of directions are
`avoided.
`
`For example the prohibited range of directions may
`[0070]
`be prevented from narrowing in response to a short term
`presence of a noise source, such as reception of reverbera-
`tions. Short term presence maybe defined as presence during
`less than 0.1 seconds.
`
`[0071] An adjustmentof the prohibited range ofdirections
`may be performed gradually in a time interval when speech
`stops after a period ofpresence of speech. For example, amay
`be gradually increased to a,,,, in a time interval of about 3
`seconds. Throughout the present specification, presence or
`absence of speech refer to the detection or non-detection of
`speech, respectively, of the system.
`[0072] Aspeech stop maybe defined as the momentthat no
`speech has been detected for e.g., 5 seconds, and a conversa-
`tion stop may be defined as the momentthat no speech has
`
`7
`
`

`

`US 2011/0103626 Al
`
`May5, 2011
`
`been detected for e.g., 30 seconds. Speech start and conver-
`sation start may be defined as the moment that speech is
`detected for the first time after a speech stop and a conversa-
`tion stop, respectively.
`[0073] A long term average maybe defined as the average
`over e.g., 2 seconds. A short term average may be defined as
`the average over e.g., 50 milliseconds.
`[0074]
`In some embodiments,
`the prohibited range of
`directions is adjusted uponstart of conversation according to
`the following:
`[0075] Calculation of the long term average DOA value
`during speech presence is performed; typically the calcula-
`tion requires 2 seconds of speech presence.
`[0076]
`Provided that the long term average DOA value
`during speech presenceis not significantly different from the
`long term average DOA value during speech absence, o is
`increased to a.,,,, With the release time, e.g. in about 3 sec-
`onds. (This situation occurs when e.g. the noise and speech
`arrive from the samedirection, in which case beamformingis
`not advantageous, or when the speaker is outside the Hall
`radius and the perceived noise field is diffuse, or when the
`SNRis low.)
`[0077]
`Provided that the long term average DOA value
`during speech presence is significantly different from the
`long term average DOA value during speech absence, the
`prohibited range of direction is adjusted in accordance with
`the following:
`[0078] When the short term average DOA value during
`speech presence remains above or around e.g. 80°, a is
`increased to G.,,,, 1n about 3 seconds. (In this case the listener
`is apparently not interested enough in the speech to turn his
`head, or heis e.g. driving a car and can not turn his headto the
`speaker.)
`[0079] When the short term average DOA value during
`speech presence does becomesignificantly lower than 80°,
`the prohibited range of directions is adjusted to just include
`the minimum ofthe short term average DOAvalue overe.g.
`the past 2 seconds, plus a safety margin of about 20° in order
`to take head movementsinto a account. This is repeated until
`speech stop. Upon speechstop, a is adjustedto e.g. o,,,,+20°,
`where ,,,,,18 equal to the maximum ofthe short term average
`DOA values measured at a speech start over e.g. the last 3
`speechstart events. (This prevents the user from missing any
`ofthe speechofinterest, while a narrow beam is also obtained
`whentheuser has focused on the speaker. A situation like this
`can occur whentheuseris in a restaurant andis alternatively
`lookingat the plate and at the person next or opposite to the
`user.)
`In the above example, preferably «a,,,, 1s 180° so
`[0080]
`that an omni-directional pattern is obtained when a is
`increasedto G,,,,, Since the omni-directional pattern imparts
`a perception to the user of being connected to the environ-
`ment.
`
` a.,,,,., equal to 90° maybe selected to maintain direc-
`[0081]
`tional suppression in the back region ofthe user.
`[0082] The prohibited range of directions may be broad-
`ened to such an extent that an existing null direction @ ends up
`residing within the prohibited range.
`[0083] According to an aspect of some ofthe embodiments,
`the signal processor is adapted to move a null direction 0
`residing within a prohibited range for a certain time period,
`e.g. 1 second, or 10 seconds, outside the prohibited range.
`This may be done momentarily or over a period oftime.
`
`[0084] A null position monitor may be provided for moni-
`toring the current null position. Whenthe current null posi-
`tion resides within the adapting prohibited range of directions
`for more than, e.g., 1 second, the signal processor moves the
`null outside the prohibited range of directions.
`[0085] An estimate of the current null position may be
`obtained by averaging the direction of arrival during adapta-
`tion. Whentherate of changeof this average is similar to the
`rate of adaptation of the null, the average will be a good
`estimate of the current null position.
`[0086] The null may be movedoutside the prohibited range
`of directions in many ways. For example, when the null
`resides within the prohibited range of directions for more
`than, e.g., 1 second, the adaptivefilter H maybe re-initialized
`so that the null is positioned outside the prohibited range of
`directions. The re-initializationfilter coefficients may be read
`from a table holding previously performed measurements or
`determinationsoffilter coefficients that position the null at,
`e.g., 0, 10, 20, .
`.
`. etc degrees. In another embodiment, the
`filter coefficients are calculated when needed.
`
`[0087] The changedposition of the null maybeselected in
`different ways. For example, the changed position may be
`selected to reside as close as possibleto its previous position,
`but outside the prohibited range of directions. In another
`embodiment, the changedposition is selected at the location
`that has the greatest distance from all prohibited ranges of
`directions that are currently in effect.
`[0088]
`For example, the adaptive filter may be forced to
`position the null direction at 8=180° and continue adaptation
`from this value. e.g., the coefficients of the adaptivefilter 52
`may be reset to values that position the null direction at
`0=180°. In another example wherein the cost function that H
`minimizes is equal to the output power, a weighted bias term
`is added to the cost function that for