throbber
PATENT NUMBER
`
`6383345
`
`6363345
`
`U.S. UTILIW PATENT APPLICATION
`
`tlAR E s ?0m
`
`ART UNIT' -,
`,4-
`;-'
`--;-',/
`
`FILEO WITH:
`
`DrsK (cRF) flrrcHe
`
`{Anached in pockst on dght insid€ tlap)
`
`PREPARED AND APPROVED FOR ISSUE
`
`z F (
`
`.)
`tl-
`
`6o
`Jo
`uJloqt
`
`tJ)
`
`,.0
`
`Ni
`
`. VN
`
`,._f
`
`t!N
`
`ORIGII{AL
`GLASS I SUBCLASS
`70q
`I aa,i'
`2t/ oA
`
`INTERNATIONAL CLASSIFICATION
`g I
`
`L
`
`tssurNG cLAsstFtcATtoN
`cRoss REF€FEI{CE(S)
`
`CLASS
`-70 c4
`
`suBcLAsSr(oNE SUBCLASS PER BLOCX)
`
`233 ,9"ar
`
`fl Continued on lssue Slip lnside File Jgckal *
`
`I lorsclarueR
`r--TTERMINAL
`
`E a) The term of this patent
`subsequent to _---._
`has been disclaimed.
`E U) fn" lerm of thls patent shall
`
`(dale)
`
`of U.S Patent. No.
`
`fl c) The terminal
`of
`lhis palenl have been disclaimed.
`-months
`
`' Sheets Drwg.
`l\r
`
`DRAWINGS
`
`Figs. Drwg.
`
`{o
`
`Print Fig.
`.")
`
`^
`
`(Assislsnt Examan6r)
`
`Ftichemond Oory,{
`PrimaryExarnifler
`
`q#l
`
`CLAIMgALLOWED
`
`Total CJFirn$
`
`Print Claim for O.G.
`
`I
`
`,&,',
`lo-t ot o/
`
`NOTICE OF ALLOWANCE MAILET'
`
`ISSUE FEE
`
`Amount Due
`
`Date Paid
`
`ff 6la, "n
`ISSUE BATCH NUMBER
`
`l)- . lt ,- ")
`
`,ulf
`
`M 52-
`
`WARNING:
`The information disclosed heroin may bo restricted. unaulhorizsd disclosure may be prohibited by tho united states code Tille 35, soctions t?2, 181 and 368,
`Possession oulside lhe U.S. Patent & Trad€mark Ottica is restrictgd to authorized employges and contractoG only.
`
`Fom PTS436A
`(Rov. 6/98)
`
`r'x){.: ) j.rjiitil S,_...
`
`q*
`
`-:{,,$,,r
`
`*0.
`(LABEL AREA)
`
`lsms For
`
`{FACE)
`
`RTL345-2_1020-0001
`
`

`
`PATENT APPLICATION
`
`0925?87tn
`
`t'
`
`-
`ic549 U.S' PTo
`09/2s28?{
`
`ililIfilry!rut[|llu|!
`
`illlllllll
`
`r1161urwrnu$-nLL-
`
`Application
`
`!- p"purt.
`
`l.
`2.
`
`3.
`
`4.'
`
`5.
`
`6.
`
`7.
`
`q
`
`9.
`a To.
`
`11.
`
`12'.
`
`17.
`
`18.
`
`19.
`
`20.
`
`21.
`
`22.
`
`23.
`
`24.
`
`25.
`
`26.
`
`27.
`
`28.
`
`29.
`
`30.
`
`31.
`
`.32.
`
`33.
`
`34.
`
`35..
`
`36.
`
`37.
`
`38.
`
`39.
`
`40.
`
`41.
`
`CONTENTS
`Date received
`(lncl. C. of M.)
`or
`Date Mailed
`
`7€.
`
`42.
`
`44.
`
`45.
`
`46.
`
`47.
`
`4 -/J- 17
`,|
`, t" tl\
`\ * id"."v
`
`ff
`
`48.
`
`s0.
`
`L(U&14, 4e.
`4 sr.
`{ ur,
`53. ,
`61 54. :
`55' ,, ,
`se. 'tl ri
`
`74.
`
`78.
`
`A*
`
`(LEFr OUTSIDE)
`
`RTL345-2_1020-0002
`
`

`
`ISSTIE SLIP STAPT ; dtEA (for arJ.-"".' '.i r;-'ss references)
`
`3-s/- rf
`
`Non-electild
`.,....., lnterference
`........ APPeal
`........0biected
`
`Dale
`
`Claim
`
`6oo
`
`u-
`
`INDEX OF CLAIMS
`......,. Reiected
`........ Allowed
`(Through numeral)... Canceled
`........ Restricted
`
`N I A 0
`
`lf more than 150 claims or 10 actions
`staple additional sheet here
`
`(LEFT INSIDE)
`
`RTL345-2_1020-0003
`
`

`
`SEARCHED
`
`SEARCH NOTES
`trNcr-uolNa SEARCH STRATEGY)
`
`JlLull
`
`j-- Zr<'o t
`
`Oa7\
`ttD
`233
`2oo
`2olZof
`ZLb
`,227
`278
`
`ZL/"'
`721
`,1 19La. e
`2-bQ
`zo t-
`ZOJ
`L75>ll
`Ltb
`21C
`r6.t2
`Lz'oz
`317.ot
`
`3L
`
`t06'o
`u46'
`cp6"t9
`u46. t t+
`+oi-of
`
`RTL345-2_1020-0004
`
`

`
`,l^'
`
`RTL345-2_1020-0005
`
`

`
`SERIAL NUMBEB
`
`09 l2E2 r874
`
`FILING DATE
`02/L8/ee
`
`CLASS
`
`381
`
`GROUP ART dNIT
`2743
`
`ATTORNEY DOCKET NO.
`
`670025-2800
`
`JoSEPII UARASH' HAIFA, ISRMLt BARuCtl BERDUGo, KTRIAT-ATA 28OOO, ISRAEL.
`
`!
`
`I
`
`:r *CONTINUINC DOI.{ESTIC DAAA* * * * * * * * * * * * * * * * * * * * *
`VERIFIED
`
`Nv< ,6M
`I
`*37 1 ( NAT, L STAGE ) DATA* * * * * * * * * * * * * * * * * rf * * *
`'t
`VERII'IED
`ilror , (lY'f
`
`n/'
`
`/
`
`,/
`,/. .
`./
`
`* :hl'OREIGN APPLICAIIONS* * * * * * t, t, t t' t, t 1/
`vERrrrED
`/
`il',@
`
`/It,
`
`TATE OR
`fouNrRY
`rtx
`
`TOTAL
`ctAfMs
`49
`
`INDEP€ISDENT
`CLAIMS
`3
`
`rF RaQurRED, FoRErcN FrtrNc LrcENsE cRANTED o)4o9/99 *t svet\,.6NTrry **
`!gf9tS!-P-ri-o[rty clcimed
`tryes [76o
`SHEETS/
`35 USC 119 (edl conditions met Eiies^EpA EMet efter Ailowance
`DRAWING
`Verified andAcknowledg*r***ffi}
`10
`THOUAS .' KO!{ALSKI
`FROUI,TSR LA?IRENCE &
`745 FIFTH AVENUE
`NEW YORK NY 10151
`
`tt)
`
`vtulEoo
`
`-rniriErE--/
`
`HAUG
`
`All Fees
`1.16 Fees {Filing)
`1.17 Fees tProcessing Ext. of time)
`1.18 Fees {lssue}
`Other
`Credit
`
`f Efft
`
`SYSTEIIT, I.ISIHOD AND APPARATUS FOR CANCELLING NoIsE
`
`utJbF
`
`FILING FEE
`FECEIVED
`
`s654
`
`FEES: Authority has been given in Paper
`No. _
`to charge/credit DEPOSIT ACCOUNT
`for the following:
`NO.
`
`RTL345-2_1020-0006
`
`

`
`F
`'1,
`
`cdi.*r-*K
`.- iuft-"
`''f-r
`*;-
`;ld
`
`FROIT{MER IJAWRENCE & HAUG tLP
`745 Fifth Avenue
`New York, New york 10151
`TeL (2L2) 5gg-0800
`Fax (2L2) 588-0500
`PATEMT APPTJTCATION TRAI{SMITTAT
`Dat.e: February 19, l?99
`Re:
`670025-2800
`TO: THE COMMISSIONER OF PATENTS AND TRADEMARKS
`Box PATENT APPLICATION
`Washington, D.C. 2O23I
`
`Sir:
`with reference to the filing
`in the unj_ted states patent ancl
`Trademark office of an application for patent j_n the name ofr
`{IOSEPH MARiilSH and BARUCH BERDUGO
`ENTitled: SYSTEM. METIIOD AI{D APPARATUS FoR CAI{CEIIJING NoIsE
`The following are enclosed:
`X Specification {22 pages) and One page of Abstract (p" j i
`49 Claims (including 3 independent -laims; pp. 23-31-)
`X
`X
`10 Sheets of Drawings.(Figs. 1, 2,3, 4,5,-5A, 6,7,
`8, 9)
`unsigned Declarati-on and power of Attorney (2 pages)
`The. f iling fee will be paid 1ater, in response t.o a
`Notice to File Missing parts. Kindly accord the
`applj-cation a February 18, 199g f iling dat.e and addres;,
`all communicat.ions to the undersigned at the address
`above.
`Respectfully submitted
`an
`
`xx
`
`Thomas
`
`Kowalski, Reg. No. 32,147
`
`EXPRESS MAII
`Mailing Label Number Etn2g9O913lUS
`DateofDeDoait @
`I hereby-cirtifyTEFtnrs paper or fee is being
`deposited rith the United SLaEes postal Service
`n&<press Mail PosC Office to Addresseen Seryice
`under 37 CFR 1.10 on the dat.e indi-caced above and
`
`Eure of person ma
`
`RTL345-2_1020-0007
`
`

`
`IN TIIE I'NITED STATES PATEIIT ATiTD TRADEMARK OFFICE
`
`APPLICATION FOR IJETTERS PATEIW
`
`Tit,Ie:
`Inventors:
`
`22 pages
`49 Claim
`-!lL sheets
`
`SYSTEITI, I,IETHOD AlrD APPARAIUS FOR CAIICELLING NOISE
`'Joseph Marash, Baruch Berdugo
`
`specification and one page of Abstract (page i)
`(3 Independent; on pages 23-3L)
`of Figs. (Figs. l--5, 5A, 6-9)
`
`EXPRESS MAII,
`
`is addresged bo the Assl"stant Comi.ssioner of
`, Washington, D.C. 20231
`
`Thomas J. Kowafski
`Registration No. 32,t47
`I. Marc Asperas
`Registration No. 37,274
`FROMMER LAWRENCE & HAUG LLP
`745 Fifth Avenue
`New York, New York 10151
`(2L2) s88-0800
`
`MARCA\2800.APP (IMA: car)
`
`RTL345-2_1020-0008
`
`

`
`PATENT
`670025-28OO
`
`RELATED APPI,ICATIONS INCORPORATED BY REFERENCE.
`The following applications and patent (s) are cited and
`hereby herein incorporated by reference: U.S. Patent Serial No.
`og/a30,923 filed August 6, 1-998, U.S. Patent serial No.
`'l , 1-998, U. S. PatenL Serial No. 09/059, 503
`Og /A55,7Ag f iled April
`filed April 13, 1998, U.S. Patent Serial No. 08/840,1-59 filed
`!4, 19g'.7, U.S. PaLent Serial No. og/I30,923 filed August 6,
`April
`L998, U.S. Patent Serial No. 08/672,899 now issued U.S. Patent
`No. 5,825,898 issued October 20, 1-998. And, al-t documents cited
`herein are incorporated herein by reference, ds are documents
`cited or referenced in documents cited herein-
`FIELD OF THE IIiIVENTION.
`The presenL invention relates to noise cancellation and
`to noise cancellation and
`reduction and, more specifically,
`reduction using spectral subtraction.
`
`BACKGROITIID OF THE INVEMTTON.
`Ambient noise added to speech degrades the performance
`of speech processing algorithms. Such processing algorithms may
`include dictation, voice activation, voice compression and oLher
`systems. In such systems, it is desired to reduce the noise and
`improve the signal to noise rat.io (S/N ratio) without effecting
`the speech and its characteristics.
`Near field noise canceling microphones provide a
`
`MARCA\2B00.APP ( IMA: car)
`
`1e:
`
`151j
`
`20
`
`25
`
`RTL345-2_1020-0009
`
`

`
`the
`
`this
`
`PATENT
`67002s -2800
`satisfactory solut.ion but require that the microphone in the
`proximity of the voice source (e.g., mouth). In many cases,
`is achieved. by mounting the microphone on a boom of a headset
`which situates the microphone at the end of a boom proximate
`mouth of the wearer. However, t.he headset has proven to be
`either uncomfortable to wear or too restricting for operation
`for example, an automobile.
`Microphone array t.echnology in general, and adaptive
`beamforming arrays in particular, handle severe directional
`noises in the most efficient way. These systems map the noise
`f iel-d and. create nul-ls towards t.he noise sources. The number of
`nulls is limited by the number of microphone elements and
`processing power. Such arrays have the benefit of hands-free
`operation without the necessity of a headset.
`However, when the noise sources are diffused, Lhe
`performance of t.he adaptive system will be reduced Lo the
`perfornance of a regular delay and sum microphone array, which is
`not always satisfactory. This is the case where t.he environment
`is quite reverberant, such as when Lhe noises are strongly'
`reflected from the wal1s of a room and reach the array from an
`infinite number of directions. Such is also the case in a car
`environment for some of t.he noises radiated from the car chassis.
`
`OBiTECTS AI{D SI'MMARY OF THE IN1TENTION
`The spectral subtraction technique provides a solution
`MARCA\2800 .APP ( IMA: car)
`
`4*
`
`rsi
`
`:1t
`?nsi
`
`.!*
`
`1si
`:lt.-
`s;c
`r&
`
`2A
`
`25
`
`RTL345-2_1020-0010
`
`

`
`PATENT
`670025-2800
`to furt.her reduce the noise by estimating the noise magnitude
`spectrum of the polluted signal. The t.echnique estimates the
`magnitude spectral level of t.he noise by measuring it during non-
`speech time intervals det.ected. by a voice switch, and then
`subt.racting the noise magnit.ude spectrum f rom t.he signal. This
`method, described in detail tn Suppression of Acoustic Noise in
`Speech Using Speetral Subtraction, (Steven F Bo7l, IEEE ASSP-27
`NO.2 Apri7, 7g7g), achieves good results for stationary diffused
`noises t.hat are not correlated with the speech signal. The
`spectral subtraction method, however, creates artifacts,
`sometimes described as musical noise, that may reduce the
`'perf ormance of t.he speech algorithm (such as vocoders or voice
`activat,ion) if the spectral subtraction is uncontrolled. In
`addit.ion, the spectral subtraction method assumes erroneously
`that. the voice switch accuralely det.ects the presence of speech
`and locates t.he non-speech t.ime intervals. This assumption is
`to achieve or
`reasonable for off-]ine systems but difficult
`obt.ain in real time systems.
`the noise magnitude spectrum is
`More particularly,
`esLimated by performing an FFT of 255 points of the non-speech
`time intervals and computing the energy of each frequency bin.
`The FFT is performed after the time domain signal is multiplied
`by a shading window (Hanning or other) with an overlap of 50?.
`The energy of eaeh frequency bin is averaged with neighboring FFT
`time frames. The number of frames is not determined but. depends
`
`MARCA\28OO.APP ( IMA : CAr)
`
`43
`
`l- q:g
`
`l-3;tr-
`
`';
`
`20
`
`25
`
`RTL345-2_1020-0011
`
`

`
`t_fff
`
`'f-i
`'.-a
`
`l atl
`
`PATENT
`670425-2800
`on the stability of the noise. For a stationary noise, it is
`preferred that many frames are averaged to obtain better noise
`estimation. For a non-stationary noise, a long averaging may be
`harmful. Problemati-ca1ly, there is no means to know a-priori
`whether the noise is stat.ionary or non-statj-onary.
`Assuming the noise magnitude spectrum estimat.ion is
`calculated, the input signal is multiplied by a shading window
`(Hanning or other), an FFT is performed (255 points or other)
`with an overlap of 50? and the magnitude of each bin is averaged
`over 2-3 FFT frames. ?he noise magini-tude spectrum is then
`subtracted from the signal magnitude. rf the result is negat.J-ve,
`the val-ue is replaced by a zero (Half Wave Rectification).
`ft is
`recommended, however, to further reduce the resi-dual noise
`present duri-ng non-speech int.ervals by replacing low values with
`a minimum value (or zero) or by at.tenuating the residual noise by
`3odB. The resurting output is the noise free magnitude spectrum.
`The spectral complex dat.a is reconstructed by applying
`the phase information of the relevant bin of the signal's FFT
`with the noise free magnitude. An IFFT process is then performed
`on the complex data to obtain t.he noise f ree time domain dat.a.
`The t.ime domain resulLs are overlapped and summed with the
`previous frame's results t.o compensate for the overlap process of
`the FFT.
`
`There are
`described. First,
`
`several problems associated with the system
`the syst.em assumes that there is a prior
`
`MARcA\2Boo -App ( rMA: car)
`
`RTL345-2_1020-0012
`
`

`
`PATENT
`67 0425- 2800
`knowledge of the speech and non-speech time intervals. A voice
`switch is not pract.ical to detect those periods. Theoretically,
`a voice swit.ch det.ects the presence of the speech by measuring
`the energy level and comparing it to a threshold. If the
`threshold is too high, there is a risk that some voice time
`intervals might be regarded as a non-speech time int.erval and the
`system will regard voice information as noise. The result is-
`voice distortion, especially in poor signal to noise ratio cases.
`Tf, on the other hand, the threshold is too low, there is a risk
`that t.he non-speech intervals will be too short especially in
`poor signal to noise raLio cases and in cases where the voice is
`continuous with little
`intermission.
`Another problem is t.hat. the magnitude cal-culation of
`the FFT result is quite complex. This j-nvol-ves sguare and square
`root calculations which are very expensive in terms of
`computation load. Yet another problem is t.he association of the
`phase information to t.he noise free magnit.ude spectrum in order
`to obtain the information for the IFFT. This process requires
`the calculation of the phase, the storage of the informat.ion, and
`applying the information to the magnitude dat.a - al-l are
`expensive in terms of computat.ion and memory requirements.
`Another problem i-s the estimation of the noise spectral
`magnitude. The FFT process is a poor and unstable estimator of
`energy. The averaging-over-time of frames contributes
`insufficiently Lo t.he stability. Shortening the length of the
`MARcA\28oo.App (rMA:car)
`
`U
`
`u
`
`Iw
`
`16J
`
`iF
`
`ri
`
`r3'i I
`
`t--
`i;f
`
`2A
`
`25
`
`RTL345-2_1020-0013
`
`

`
`PATENT
`670025 -2800
`FFT results in a wider bandwidth of each bin and better stability
`but reduces the performance of the system. Averaging-over-time,
`moreover, smears the data and, for t.his reason, cannot be
`extended to more than a few frames. This means that the noise
`estimat.ion process proposed is not sufficienLly stable.
`It is t.herefore an object of this invention to provide
`a spectral subtraction system that has a simple,.YeL efficient
`mechanism, to estimate the noise magnitude spectrum even in poor
`signal-to-noise ratio situations and in continuous fast speech
`cases.
`
`It is another object of this invent.ion to provide an
`efficient mechanism that. can perform the magnitude estimation
`with littLe cost, and will overcome the probfem of phase
`association.
`It is yet another object of this invention Lo provide a
`stable mechanism to estimate t.he noisei spectral magnitude wit.hout
`the smearing of the dat.a
`In accordance wit.h t.he foregoing objectives, the
`present invention provides a system that correctly determines the
`non-speech segments of t.he audio signal t.hereby preventing
`erroneous processing of the noise canceling signal during the
`speech segments. In the preferred embodiment, the present
`invention obviates the need for a voice switch by precisely
`d.etermining t.he non-speech segments using a separate threshold
`detector for each frequency bin. The threshol-d detector
`MARcA\2Boo.App (fMA:car)
`
`,4 6
`tl
`I
`
`i"c
`r_sf
`
`r3d
`qgi
`
`rfrts
`
`irE
`
`151,;
`i-r
`
`. i|h
`
`20
`
`25
`
`RTL345-2_1020-0014
`
`

`
`PATENT
`67A025 -2800
`precisely detect.s the positions of t.he noise elements, even
`within continuous speech segments, by det.ermining whether
`frequency spectrum elements, ,or bins, of t.he input signal are
`within a threshold set according to a'minimum value of the
`freguency spectrum efements over a preset period of time. More
`precisely, current and future minimum values of the frequency
`spectrum elements. Thus, for each syllable, the energy of the
`noise elements is det.ermined by a separate threshol-d
`determi-nation without examination of the overall signal energy
`thereby providing good and st.able estimat.ion of the noise. In
`addition, the slrstem preferably sets t.he t.hreshold continuously
`and resets the threshold within a predetermined period of time
`of, for example, five seconds.
`In order to reduce complex caleulations, iL is
`preferred in the present invenLion to obtain an estimate of the
`magnitude of the input. audio signal using a multiplying
`combination of the real and imaginary parts'of the input in
`accordance with, for example, the higher and the lower values of
`In ord"er to further
`the real and imaginary parts of the signal.
`reduce instability of the spectral estimation, a two-dimensional
`(2D) smoothing irocess is applied to the signal estimation. A
`two-step smoothing function using first neighboring frequency
`bins in each time frame then applying an exponent.ial- t.ime average
`effecting an average over t.ime for each freguency bin produces
`excellent result.s.
`MARCA\28o0.APP (rMA:car)
`
`7
`
`4
`U
`
`1df
`
`iiT
`
`.";
`
`;F
`
`l.3:f'
`
`20
`
`RTL345-2_1020-0015
`
`

`
`t**
`I V;ot
`
`i:5
`i* i'
`
`,Es
`
`!'"n
`
`l_5id
`
`PATENT
`67A025-2800
`In order to reduce Lhe complexity of determining the
`phase of t.he frequency bins during subtraction to thereby align
`the phases of t.he subtracting elements, the present invention
`applies a filt.er multiplication t.o effect the subtraetion. The
`function, a Weiner fil-ter function for example, or an
`filter
`is multiplied by the complex
`approximation of the Weiner filter
`dat.a of the frequency domain audio signal. The filter
`function
`may effect a, fulI-wave rectification, ot a half-wave
`rectifi-cation for otherwise negat.ive results of the subtraction
`process or simple subtraction. It will be appreciated that,
`since the noise elements are determined within conLinuous speech
`segments, the noise estimation is accurate and it. may be canceled
`from t.he audio signal conti.nuously providing excellent noise
`cancellation characteristics .
`The present invention also provides a residual noise
`reduction process for reducing the residual noise remaining afLer
`noise cancellation. The residual noise is reduced by zeroing the
`t.he continuous speech, or
`non-speech segments, €.g., within
`decaying the non-speech segments. A voice switch may be used or
`another threshold detector which detects the non-speech segments
`in the time-domain.
`The present invention is applicable with various noise
`canceling systems including, bu1 not limit.ed to, those systems
`described in the U.S. patent applications incorporated herein by
`reference. The present invenLion, for exampJ-e, is applicable
`MARCA\2g00.APP (IMA: car)
`
`8
`
`Cl
`
`i
`
`RTL345-2_1020-0016
`
`

`
`10:g
`r{#
`
`iqi
`
`!F
`
`. df,;
`I ?:.I
`i'-
`
`PATENT
`670025 -2800
`wit.h the adapt.ive beamforming array.
`In addition, the present
`invention may be embodied as a computer program for driving a
`computer processor either instal-led as application software or as
`hardware.
`
`BRIEF DESCRIPTION OF THE DRAWINGS
`Other objects, features and advantages according to the
`present invent.ion will become apparent from the following
`detailed description of the illustrated embodiments when read in
`conjunction with the accompanying drawings in which corresponding
`components are identified by the same reference numerafs.
`Fig. l- illustrates
`the present invent.ion;
`Fig. 2 il-lustrates the noise processing of the present
`invention;
`Fig. 3 illustrates
`the noise estimaLion processing of
`t.he present. invention;
`Fig. 4 itlustrates t.he subtraction processing of the
`presenL invention;
`Fig. 5 illustrates
`present invention;
`Fig. 5A illustraLes a variant of Lhe residual noise
`processing of the present invention;
`Fig. 5 illustrates a flow diagram of the present
`invention;
`Fig. 7 illustrates a fl-ow diagram of Lhe presenL
`
`the residual noise processing of the
`
`MARcA\2goo.APP ( rMA: car)
`
`f8
`
`RTL345-2_1020-0017
`
`

`
`PATENT
`57A025-2800
`
`invention;
`Fig. 8 illustraLes a flow diagram of the present
`invention; and
`Fig. 9 illustrates a flow diagram'of t.he present
`'invention.
`
`DETAIIJED DESCRIPTION OF' THE PREFERRED EMBODIME}ffTS
`Figure 1 illustraLes an embodiment. of the present
`invention l-00. The system receives a digital audio signal at
`input aA2 sampled at a frequency which is at least twice the
`In one embodiment, the signal is
`bandwidth of the audio signal.
`derived from a microphone signal t.hat has been processed through
`an analog fronl end, A/D converter and a decimation filter
`to
`obtain the required sampling freqr.rency. In another embodiment,
`the input. is t,aken from the output. of a beamformer or even an
`adaptive beamformer. In that case the signal has been processed
`to eliminate noises arriving from directions other than the
`desired one leaving mainly noises originated from the same
`direction of the desired one. fn yet another embodiment, Lhe
`input signal can be obtained from a sound board when the
`processing is implemented on a PC processor or similar computer
`processor.
`
`1d.I
`
`lori
`
`o'
`
`lEij
`
`irq?t$
`
`-iq
`
`20
`
`The input samples are
`of 255 point.s. When the buf fer
`combined in a combiner 105 with
`
`stored in a temporary buffer 104
`the new 255 points are
`is full,
`the previous 255 points to
`
`25
`
`MARCA\28OO.APP (IMA: CAr)
`
`1-0
`
`RTL345-2_1020-0018
`
`

`
`PATENT
`570025-2800
`provide 512 input point.s. The 51-2 input points are mul-Liplied by
`multiplier 108 with a shading window with the length of 5L2
`points. The shading window contains coefficients that are
`multiplied with the input data accordj-ngly. The shading window
`can be Hanning or othe'r and it serves two goals: the f irst is to
`smooth t.he transients between two processed blocks (together with
`t.he overlap process) ; t.he second is to reduce the side lobes in
`the frequency domain and hence prevenL the masking of 1ow energy
`tonals by high energy side lobes. The shaded results are
`converted t.o t.he frequency domain through an FFT (Fast Fourier
`Transform) processor 110. Other lengths of the FFT samples (and
`accordingly input buffers) are possibte including 256 points or
`AO24 points
`The FFT output is a complex vector of 256 significant
`points (the other 256 points are an anti-symmet.ric replica of Lhe
`first 256 points). The points are processed in the noise
`processing block LL2i€AO) which includes the noise magnitude
`estimation for each frequency bin - the subtraction process that
`estimates the noise-free complex value for each frequency bin and
`the residual noise reduction process. An IFFT (Inverse Fast
`Fourier Transform) processor 1L4 performs the Inverse Fourier
`Transform on the complex noise free data to provide 5L2 time
`domain points. The fi::st 256 time domain points are summed by
`the summer 115 with the previous last 256 data points to
`compensate for the input overlap and shading process and output
`MARcA\28oo.APP (rMA:car)
`
`11
`
`l,v
`
`rlFd
`
`tit
`
`20
`
`25
`
`RTL345-2_1020-0019
`
`

`
`PATENT
`570025 -2800
`at output terminal 118. The remaining 255 points are saved for
`t.he next iteration.
`It will be appreciated that, while specific transforms
`are utilized in the preferred embodimenLs, it is of course
`understood t.hat ot.her transforms niay be applied to the present
`invention to obtain the spectral noise signal.
`Figure 2 is a detailed description of the noise
`processing block 2OO(11'2). First, each frequency bin (n) 202
`magnitude is estimated. The straight forward approach is to
`i=t1-d:: est.imate t.he magnitude by calculating:
`
`5
`
`riT
`
`il$ v h) = ( (Rear (n) )2+ (rmas (n) )2 )-2
`.i[
`*
`In order to save processing t.ime and complexity the
`1g]j signal magnit.ude (Y) is estimated by an estimator 2o4 using an
`{i} approximation formula instead:
`
`lid
`
`20
`
`25
`
`y (n) = Max [lneal (n) ,Imag (n) | ] +A .4* Min tlReaI (n) ,Imag (n) | l
`
`In order to reduce the instability of the spectral
`estimation, which typically plagues the FFT Process (ref [2]
`DigitaT Signal Processing, Oppenheim Schafer, Prentice HaLl P:
`542545), the present invention implements a 2D smoothing process.
`Each bin is replaced with the average of it.s value and the two
`neighboring bins' value (of the same time frame) by a first
`MARCA\28O0.APP (IMA:car)
`12
`
`i"l
`I5
`
`RTL345-2_1020-0020
`
`

`
`PATENT
`670025-28A4
`averager 206. In addition, the smoothed value of each smoothed
`bin is further smoothed by a second averager 208 using a t.ime
`exponential average with a time constant of A.7 (which is the
`equivalent of averaging over 3 time frames). The 2D-smoothed
`value is then used by two processes - Lhe noide estimation
`process by noise estimati-on processor 2L2(300) and the
`subt.raction process by subtract or 2IO. The noise estimation
`process estimaLes the noise at each frequency bin and the result
`is used by the noise subtraction process. The output. of the
`noise subtraction is fed inLo a residual noise reduction
`processor 2ir6 to further reduce the noise. In one embodiment,
`the time domain signal is also used by the residual noise process
`2:-6 to determine the speech free segments. The noise free signal
`is moved to the IFFT process |o obtain t.he time domain output
`2L8.
`
`Figure 3 is a det.aifed description of the noise
`estimation processor 3OO (2L2) - Theoretieally, the noise should
`be estimated by taking a long time average of the signal
`tiris requires that a
`magnit.ude ivl of non-speech time intervals.
`voice switch be used to detect. the speech/non-speech intervals.
`However, a Loo-sensiLive a switch may result in the use of a
`speech signal for the noise est.imation which will defect the
`voice signal. A less sensitive switch, on the other hand, may
`dramatically reduce the length of the noise time int.ervals
`(especially in continuous speech cases) and defect t.he validity
`
`5
`
`1d:n
`
`nqF
`
`IJ'F
`
`20
`
`25
`
`MARCA\2800.APP (IMA: car)
`
`*13
`
`RTL345-2_1020-0021
`
`

`
`PATENT
`67 0025-2804
`
`r
`
`rJ-
`
`of the noise estimation.
`In the present invention, a separat.e adapt.ive threshold'
`is implemented for each frequency bin 302. This alfows the
`location of noise elemenLs for each bin separately without the
`examination of the overall signal energy. The logic behind this
`method is that, for each syl1ab1e, the energy may appear at I
`different. frequency bands. At Lhe same time, oLher frequency
`bands may contain noise elements. It is therefore possible to
`apply a non-sensitive threshold for the noise and yet locate many
`10:[ non-speech data points for each bin, even within a continuous
`i*q
`uf
`speecn case. The advantage of this method is that it allows the
`ii:
`;;{
`tiI
`il+ collect.ion of many noise segments for a good and sLable
`tr,* esLimation of the noise, even within continuous speech segments'
`In t.he threshold determination process, for each
`ir
`frequency bin, two minimum values are calculated. A future
`lgii
`!-*i:-hf minimum value is initiated every 5 seconds at 304 with the value
`,ig of the current magnitude (Y (n) ) and replaced with a smaller
`UF
`.--.
`..
`1 --rLL
`-
`minimal value over the next 5 seconds through the following
`process. The future minimum value of each bin is compared with
`If the currenL
`the current magnitude value of the signal.
`magnitude is smaller t.han the future minimum, the future minimum
`is replaced with 1he magnitude which becomes the new future
`minimum-
`
`5
`
`2A
`
`25
`
`At. the same time, a current minimum value is calculated
`at 306. The current. minimum is iniLiated every 5 seconds with
`MARCA\28oo.APP (rMA:car)
`. Un
`IL
`t .'r
`
`RTL345-2_1020-0022
`
`

`
`PATENT
`670025 -2800
`the value of the future minimum that was determined over the
`previous 5 seconds and follows the minimum value of t.he signal
`for the next 5 seconds by comparing its value with the current
`magnitude value. The current minimum value is used by the
`subtraction process, while the future minimum is used for t.he
`initiatj-on and refreshing of the current minimum.
`The noise estimation mechanism of the present invention
`ensures a t.ight and quick estimation of the noise va1ue, with
`limited memory of the process (s seconds), while preventing a too
`high an estimation of the noise.
`Each bin's magnitude (Y(n) ) is compared with four times
`the current minimum value of that. bin by comparator 308 - which
`If the magnitude
`serves as the adapt.ive t.hreshold for that bin.
`is within the range (hence bel-ow the threshol-d) , it is al-lowed as
`noise and used by an exponential averaging unit 310 that
`det.ermines the level of the noise 312 of that frequency. If the
`magnitude is above the t.hreshold it is rejected for t.he noise
`estimation. The t.ime constant for the exponential averaging is
`typically 0.95 which may be interpreted as taking the average of
`the last 20 frames. The threshold of *minimum val-ue may be
`changed for some applications.
`Figure 4 is a detailed description of the subt.raction
`In a straight forward approach, Lhe value of
`processor 400 (21-0) .
`the estimated bin noise magnitude is subtracted from the current
`bin magnitude. The phase of the current. bin is calculated and
`15
`
`MARcA\2Boo.APP (rMA: car)
`
`I'v
`
`rl.#
`
`";-
`
`ni
`
`20
`
`25
`
`RTL345-2_1020-0023
`
`

`
`PATENT
`670025-2800
`used in conjunction with the result of t.he subtraction Lo obtain
`the Real and Imaginary parts of the resuft. This approach is
`very expensive in terms of processing and memory beeause it
`requires the calculation of the Sine and Cosine arguments of the
`complex vecLor with consideration of the 4'quarters where the
`complex vector may be posit.ioned. An alternative approach used
`in this present invention is Lo use a Filter approach. The
`subtraction is interpreted as a filter multiplicaLion performed
`by filter 402 where H (the filter coefficient) is:
`
`H(n) =
`
`llY(n)l - lrv,n)ll
`, lvtn)l
`Where y(n) is the magnitude of the current bin and N(n)
`is the noise estimation of t.hat bin. The val-ue H of the fi-lter
`coefficient (of each bin separately) is multiplied by the Real
`and Imaginary parts of the current bin at 4042
`
`E (Real) =Y (Real) *H
`
`;
`
`E (Inag) =Y (Imag) *H
`
`Where E is the noise free complex value. In the
`straight. forward approach t.he subtraction may result in a
`negative value of magnitude. This value can be either replaced
`with zero (hal-f-wave rectification) or replaced with a positive
`value equal to t.he negative one (full-wave rectification).
`The
`filt.er approach, as expressed here, results in the full-wave
`L6
`
`MARCA\28OO.APP ( IMA: CAr)
`
`#1
`
`i-O:t'
`
`-;!T/D l;;i
`lnll iliI lsr
`r)
`.-
`
`i.cl1f4
`tr-i
`
`-.i:
`15;
`gn
`
`t--
`irF
`r t:*
`ii3
`
`20
`
`25
`
`RTL345-2_1020-0024
`
`

`
`PATENT
`570A25-2804
`rectification directly. The full wave rect.ification provides a
`less noise reduction but introd-uces much less artifacts to
`little
`It will be appreciated Lhat this filter can be
`the signal.
`modified to effect a half-wave rectification by taking the non-
`absolute value of the numerator and replacing negative values
`wit.h zeros.
`Note also that the values of Y in the figures are the
`smoothed values of Y after averaging over neighboring spectral
`bins and over time frames (2D smoothing). Another approach is to
`use the smoothed Y only for the noise estimation (N), and Lo use
`the unsmoothed Y for the cal-culation of H.
`t.he residual noise reduction
`Figure 5 illustrales
`processor 50 O Q!6) . The residual noise is defined as the
`remaining noise during non-speech intervals. The noise in these
`reduced by the subtraction process which does
`int.ervals is first
`not differentiate between speech and non-speech time intervals.
`The remaining resid.ual noise can be reduced further by using a
`voice switch 502 and either multiplying the residual noise by a
`decaying factor or replacing it. wit.h zeros. Another alternative
`to the zeroing is replacing the residual noise with a minimum
`value of noise at 504.
`Yet another approach, which avoids t.he voice switch, is
`illust.rated in Figure 5A. The residual noise reduction processor
`506 appties a similar threshold used by the noise estimator at
`508 on the noise free output bin and replaces or decays the
`MARcA\2goo.APP (rMA:car)
`17
`
`ai/\IJ
`
`1_ 0"r
`
`'";
`
`1Ei-i
`
`- li.q
`
`20
`
`25
`
`RTL345-2_1020-0025
`
`

`
`1q:;
`
`trc
`
`!Ei{F
`
`15i#
`
`PATENT
`67002s -280.0
`
`of the
`intervals.
`noise when
`and Lhe
`
`result when it is lower than t.he threshold at 510.
`'
`The result of the residual noise processing
`present invention is a quieter sound in Lhe non-speech
`However, the appearance of art.ifacts such as a pumping
`the noise level is switched between the speech interval
`non-speech interval may occur in some applications.
`The spectral subtracLion technique of Lhe present
`invention can be utilized in conjunction with the array
`technigues, close talk microphone technique or as a stand alone
`system. The spectral subtraction of the present invention can be
`implemented.on an embedded hardware (DSP) as a stand alone
`system, as part of other embedded algorithms such as adaptive
`beamforming, or as a software application running on a PC using
`data obtained from a sound port.
`As illust.rat.ed in Figures 6-9, for example, the present
`In step
`invention may be implemented as a software application.
`600, the input samples are read. At step 602, the read

This document is available on Docket Alarm but you must sign up to view it.


Or .

Accessing this document will incur an additional charge of $.

After purchase, you can access this document again without charge.

Accept $ Charge
throbber

Still Working On It

This document is taking longer than usual to download. This can happen if we need to contact the court directly to obtain the document and their servers are running slowly.

Give it another minute or two to complete, and then try the refresh button.

throbber

A few More Minutes ... Still Working

It can take up to 5 minutes for us to download a document if the court servers are running slowly.

Thank you for your continued patience.

This document could not be displayed.

We could not find this document within its docket. Please go back to the docket page and check the link. If that does not work, go back to the docket and refresh it to pull the newest information.

Your account does not support viewing this document.

You need a Paid Account to view this document. Click here to change your account type.

Your account does not support viewing this document.

Set your membership status to view this document.

With a Docket Alarm membership, you'll get a whole lot more, including:

  • Up-to-date information for this case.
  • Email alerts whenever there is an update.
  • Full text search for other cases.
  • Get email alerts whenever a new case matches your search.

Become a Member

One Moment Please

The filing “” is large (MB) and is being downloaded.

Please refresh this page in a few minutes to see if the filing has been downloaded. The filing will also be emailed to you when the download completes.

Your document is on its way!

If you do not receive the document in five minutes, contact support at support@docketalarm.com.

Sealed Document

We are unable to display this document, it may be under a court ordered seal.

If you have proper credentials to access the file, you may proceed directly to the court's system using your government issued username and password.


Access Government Site

We are redirecting you
to a mobile optimized page.





Document Unreadable or Corrupt

Refresh this Document
Go to the Docket

We are unable to display this document.

Refresh this Document
Go to the Docket