CA1308362C - Noise suppression system - Google Patents
Noise suppression systemInfo
- Publication number
- CA1308362C CA1308362C CA000612604A CA612604A CA1308362C CA 1308362 C CA1308362 C CA 1308362C CA 000612604 A CA000612604 A CA 000612604A CA 612604 A CA612604 A CA 612604A CA 1308362 C CA1308362 C CA 1308362C
- Authority
- CA
- Canada
- Prior art keywords
- snr
- noise
- channel
- gain
- energy
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Lifetime
Links
Landscapes
- Noise Elimination (AREA)
Abstract
ABSTRACT OF THE DISCLOSURE
An improved noise suppression system (800) is disclosed which performs speech quality enhancement upon the speech-plus-noise signal available at the input (205) to generate a clean speech signal at the output (265) by spectral gain modification. The improvements of the present invention include the addition of a signal-to-noise ratio (SNR) threshold mechanism (830) to reduce background noise flutter by offsetting the gain rise of the gain tables until a certain SNR threshold is reached, the use of a voice metric calculator (810) to produce more accurate background noise estimates via performing the update decision based on the overall voice-like characteristics in the channels and the time interval since the last update, and the use of a channel SNR
modifier (820) to provide immunity to narrowband noise bursts through modification of the SNR estimates based on the voice metric calculation and the channel energies.
An improved noise suppression system (800) is disclosed which performs speech quality enhancement upon the speech-plus-noise signal available at the input (205) to generate a clean speech signal at the output (265) by spectral gain modification. The improvements of the present invention include the addition of a signal-to-noise ratio (SNR) threshold mechanism (830) to reduce background noise flutter by offsetting the gain rise of the gain tables until a certain SNR threshold is reached, the use of a voice metric calculator (810) to produce more accurate background noise estimates via performing the update decision based on the overall voice-like characteristics in the channels and the time interval since the last update, and the use of a channel SNR
modifier (820) to provide immunity to narrowband noise bursts through modification of the SNR estimates based on the voice metric calculation and the channel energies.
Description
3~
IMPROVED NOISE~ SUPPR~SION SYS~M
Cross-Referenoes to Related Applica~ons This application contains subject matter related to U.S. Patent No.
4,630,304 alld U.S. Patent No. 4,630,305, assigned to the same assignee as the present application.
Backgrolmd of the Invention 1. Field of the Invention The present invention relates generally ~o acoustic noise suppression systems. The present invention is more specifically directed to improving the speech quality of a noise suppression system employing the spectral subtraction noise suppression technique.
1~
IMPROVED NOISE~ SUPPR~SION SYS~M
Cross-Referenoes to Related Applica~ons This application contains subject matter related to U.S. Patent No.
4,630,304 alld U.S. Patent No. 4,630,305, assigned to the same assignee as the present application.
Backgrolmd of the Invention 1. Field of the Invention The present invention relates generally ~o acoustic noise suppression systems. The present invention is more specifically directed to improving the speech quality of a noise suppression system employing the spectral subtraction noise suppression technique.
1~
2. Description of the Prior Art Acoustic noise suppression in a speech communication system generally serves the purpose of improving the overall quality of the desired audio signal by filtering environrnental background noise from the desired speech signal. This ~0 speech enhancement process is particularly necessary in environments having abnorrnally high levels of ambient background noise, such as an aircraft9 a moving vehicle, or a noisy factory.
The noise suppression technique described in the aforementioned patents is the spectral subtraction -- or spectral gain modification -- technigue. Using this ~5 approach, the audio input signal is divided into individual spectral bands by a bank of bandpass filters, and particular spectral bands are attenuated accordingto their noise energy content. A spectral subtraction noise suppression prefilter utilizes an estimate of the 13~;2 " CM00370H
background noise power spectral density ko generate a signal-to-noise ratio (SNR) of the speech in each channel, which, in turn, is used to compute a gain factor for each individual channel. The gain factor is used as 05 a pointer for a look-up table to determine th~
attenuation for that particular ~pectral band. The channels are then attenuated and recombined to produce the noise-suppressed output waveform.
In speGialized applications involving rPlatively 10 high background noise environments, most noise suppression techniques exhibit signi~icant performance limitations. One exampla of uch an application is the vehicle speakerphone option to a cellular mobile radio telephone system, which provides hands-free operation ~or 15 the automobilQ driver. The mobile hands free microphona is typically located at a greater distance from the user, such a~ being mounted overhead on the visor. The more distant microphone delivers a much poorer signal-to-noise ratio to the land-end party due to road and wind noise 20 conditions~ Although the received speech at the land-end is usually intalligible, continuous exposure to such background noise leveIs often increases listener ~atigue.
Although ~ost prior art techniques perform sufficiently well under nominal background noise 25 conditions, the per~ormanc~ of known techniqu2~ becomes severely limited in such specializ~d applications of unusually high background noise. Typical spectral subtraction noise suppre~sion sy tems may reduce the background noise level over the voice ~requency spectrum 30 by as much as 10 dB without seriou~ly a~fecting the speech quality. However, when the prior art tPchniques are used in relatively high background noise environments requiring noise suppression level~ approachtng 20 d~, there is a substantial degradation in the quality 35 characteristics of the voice. Furthermcre, in rapidly-changing high noise environmant~, a sevQre low ~requency r r CM0037OH
noise flutter develops in the output speech siynal which resembles a distant "jet engine roar" sound. This noise flutter is inherent in a spectral subtraction noise suppression system, since the individual channel gain 05 parameters axe continuously being updated in response to the changing background noise environment.
The bacXground noise flutter problem was indirectly addressed but not eliminated through the use of gain smoothing. For example, R.J. McAulay and ~.L.
10 Malpas~, in tha article entitled "Speech Enhancement Using a Soft-Decision Noise SuppresRion Filter", IEEE
Tran~O Acoust., Speech, Siqnal Proces ing, Vol. ASSP-28, No. 2 (April 1980), pp. 137-145, propose the use of gain smoothing on a per-frama basis to avoid the introduction 15 of discontinuitie~ in the output waveform. Since the introduction of gain smoothing can cause the noise suppre~ion prefilter to be slow to respond to a leading edge tran ition (which would result in speech distortion), a weighting ~ac~or of 1 or 1/2 was chosen 20 such that the prefilter responds immediately to an increas~ in gain while tending to smooth any decrease in gain. Unfortunately, excessive gain smoothing still produce~ noticeable detrimental effect~ in voice quality, the primary effect being the apparent introduction of a 25 tail-end ~cho or "noi~e pump" to spoken words. There is also a significan~ reduction in voice amplitude with large amoun~ of gain smoothing.
The noise flutter performance was further improved by th~ technigue of smoothing the noise 30 suppression gain fac~ors for each individual channel on a per~sample basis instead of on a per-frame basis. Per-sampl~ ~moothing, a~ well a~ utilizing different smoothing coef~icient~ for each channel, is described in U.S. Patent No. 4,630/305j entitled "Automatic Gain 35 Selector for a Noise Suppres~ion System." However, none Or the known prior art techniques appreciate that the ~ 3 --rr CM00370H
primary source of the channel gain discontinuities is the inherent fluctuation of background noise in each channel from one frame to the next. In known spectral subtraction syst~ms, even a 2 dB SNR variation would 05 create a few dB o~ gain variation, which is then heard as an annoying background noiss flutter. Hence, the flutter problem has never been effectively sol~ed.
Moreover, narrowband noise -- that which has a high power spectral density in only a ~ew channel~ --10 further complicates the background noise ~lutter problem.~i~ce these ~ew hlgh energy noi~e channels would not be attenuated by the background noise suppressor, the resultan~ audio output has a "running water" type of charac~eristic. Narrowband noise bursts also degrade the 15 accuracy of the background noise update decision required to perform noise ~uppression in changing background noise environment Since the gain factors are chosen by SNR
estimates, which are determined by the speech energy in 20 each channel (signal) and the current background noise energy estimate in each channel (noise), the performance of the entir~ nois~ suppression system i~ based upon the accuracy of the background noise estimate. The statistic~ o~ the background noise are estimated during 25 the time when only background noise is present, such as during the pauses in human speech. There~ore, an accurate speech/noise clas i~ication must be made to determine when such pause~ in speech are occurring.
It is widely known that the energy histogram 30 technigue for distinguishing between background noise and ~paech porform suf~iciently well in normal ambient noise environment~. See, Q.g., W.J. Hes~, "A Pitch Synchronous Digital Feature Extraction System for Phonemic Recognition of Speech," IEEE Trans. Ac u~ Spe_ h, 35 Siqn~l Processin~, Vol. ASSP-24, No. i (February 1976), pp. 14-25. Energy hi~togram~ o~ acou8tic ~ignal8 exhibit r' CMo0370H
a bimodal distribution in which the two modes correspond to noise and speech. Thus, an appropriate threshold can be set between the two modes to provide the speech/noise classi~ication. However, the dis~inction between 05 background noise enargy and unvoiced speech snergy in relatively high background noise environments is wlclear.
Consequently, the task of accurately finding the two modes o~ the energy histogram, and setting the appropriate threshold between them, is extremely 10 dif~icult.
To accommodate changing noi~s backgrounds, McAulay and Malpass implement an adaptive threshold by constantly monitoring the his~ogram energy on a frame-by frama basis, and updating the threshold utilizing 15 different decay factors. Alternatively, U.S. Patent No.
4,630,304 utilizes an energy valley detector to perform the speech/noise decision based upon the post-processed signal energy -- signal en~rgy available at the output of the noise suppression system ~- to determine ~he detected 20 speech minimum. Thus, the accuracy o~ the background noise estimate is improved since it is based upon a much cleansr spe~ch signal.
Howev~r, n~ith~r prior art technique is properly responsive to a sudden, strong increase in background 25 noisQ level. ThesQ background noise estimate updating decision processes interpret a sudden, loud noise level rise as speech, such that no updates are performed. The en~rgy histogram or valley detector has a slow adaptation characteristic which will evantually adapt to the higher 30 noise level. However, this adaptation characteristic does lead to incorrect nois~ updates on the weaker energy portions of speech. Thi~ erroneous decision significantly degrades the performance of t~e noise suppres~ion system.
A ne~d, therefore exists for an improved acoustic noisQ suppression system which addr~sses the problem~ o~
B~
backgrolmd noise fluctuation, narrowband noise bursts, and sudden background noise increases.
Summsly of the Invention Accordingly, it is an object s)f the present invention to provide an improved method and apparatus for suppressing background noise in high background noise environments without significantly degrading the voice quality.
Another object of the present invention is to provide an improved noise suppression system that addresses the background noise fluctuation problemwithout requiring large amounts of gain smoothing.
A further object of the present invention is to provide a spectral 15 subtraction noise suppression system which compensates for the detrimental effects of narrowband noise bursts.
Another object of the present invention is to provide a background noise estimation mechanism which is not misled by low energy portions of speech,20 yet still provides correction for sudden, strong increases in background noise levels.
These and other objects are achieved by the present invention which, briefly described, is an improved noise suppression system for attenuating 25 the background noise from a noisy input signal to produce a ns)ise-suppressedoutput signal by spectral gain modification. The noise suppression system 800 includes a mechanism 210 for separating the input signal into a plurality of pre-processed signals representative of selected frequen~y channels, a mechanism 310for generating an estimate of the signal-to-noise ratio (SNR) in each individual30 channel; a mechanism S90 for producing a gain value for each individual channel by automatically selecting one of a plurality of gain values from a particular gain table in i ~
.
response to the channel SNR estimates, and a mechanism 250 for modi~ing the gain of each of the pluralit~y of pre-processed signals in response to ~he selected gain values to provide a plurality of post-processed noise-suppressed output 5 signals. The improvements of the present invention relate to the addition of an SNR threshold mechanism 830 to eliminate minor gain fluctuations for low SNR
conditions, a voice metric calculator 810 to produce a more accurate background noise estimate update decision, and a channel SNR modifier 820 to suppress narrowband noise bursts.
More specifically, the first aspect of the present invention pertains to the addition of an SNR threshold mechanism 830 for providing a predetermined SNR threshold which the channel SNR estimates must exceed before a gain value above a predefined minimum gain value can be produced.
I~ In the preferred embodiment, the SNR threshold is set at 2.25 dB SNR, such that minor background noise fluctuations do not create step discontinuities in the noise suppression gams.
According to the second aspect of the present invention, a voice 20 metric calculator 810 is utilized to perform the speech/noise classification for the background noise update decision using a two-step process. First, the raw SNR
estimates are used to index a voice metric table to obtain voice metric values for each channel. A voice metric is a measurement of the overall voice-like characteristics of the channel energy. The individual channel voice metric values are summed to create a first multi-channel energy parameter, and then compared to a background noise update threshold. If the voice metric sum does not meet the thresllold, the input frame is deemed to be noise, and a background noise update is performed. Secondly, the time since the occurrence of the previous background estimate update is constantly monitored. If too much time has passed 30 since the last update, e.g., 1 second, then it is assumed that a substantial increase in noise has occurred, and a background noise update is per~ormed regardless of 5 whether it looks like a voice frame. This second test is based on the assumption that speech seldom contains continuous high energy levels in all channels for more than one second, which would be the case for a sudden, loud noise level increase. The voice metric algorithm incorporating the two-step decision processprovides a very accurate background noise estimate update signal.
In the third aspect of the present invention, a channel SNR
modifying mechanism 820 provides a second multi-channel energy parameter in response to the number of upper-charmel SNR estimates which exceed a predetermined energy threshold, e.g., 6 dB SNR. If only a few channels have an 15 energy level above this energy threshold (such as would be the case for a narrowband noise burst), the measured SNR for those particular channels would be reduced. Moreover, if the aforementioned voice metlic sum is less than a metric threshold (which would indicate that the frame was noise), all channels are similarly reduced. This SNR modifying technique is based on the assumption that 20 typical speech exhibits a majority of channels having signal-to-noise ratios of 6 dB
or greater.
Brief Description of the Drauings The features of the present invention which are believed to be novel are set forth with particularity in the appended claims. The invention itself, however, together with further objects and advantages thereof? may best be understood by reference to the following description when taken in conjunction with the accompanying drawings, in which:
Figure 1 is a detailed block diagram illustrating the preferred embodiment of the improved noise .
suppression system according to the present inYent;on;
Figure 2 is a graph representing voice me~ric values output as a func~ion S of SNR estimate index values input for the voice metric calculator block of Figure l;
Figare 3 is a representative gain table graph illustrating the overa~l channel attenuation for particular groups of channels as a function of the SNE~ estima$e;
and Figures 4a through 4f are flowcharts iliustrating the specific sequence of operations performed in accordance with the practice of the preferred embodiment of the present invention.
Detailed Description of the Preferred Ernbodiment Figure 1 is a detailed block diagram of the preferred embodiment of the present invention. ~1 the elements of Figure 1 having re~erence numerals less than 600 correspond to those of U.S. Patent No. 4~628,529 - Borth et al. Refer to the Borth patent for their descriptioll. Ihe additional circuit components having referense numerals greater than 600 represent the improvements to the 20 system, and will be described herein.
Improved noise suppression 800 incorporates changes to the aforementioned Borth noise suppression system in three basic areas: (a) the updating of background noise estimates by voice metric calculator 810; (b) the modification of SNR estimates by channel SNR modifier 820; and (c) utilization 25 of SNR threshold block 830 to offset the gain rise of each channel Each of these improvements will be described in terms of the block diagram of Figure 1, and in terms of the flowchart of Figure 4a - 4f.
Voice metric calculator 810 replaces the valley detector circuitry of the previous system. A voice metric is essentially a measurement of the overall voice-like characteristics o~ the channel ~nergy. Inkhe preferred e~bodiment, voice metric calculator 810 is implemented as a look up table which translates the individual channel SNR estimate~ at 235 into voice metric values. The VoiCQ metric values are used internally to S determins when to update the bacXground noise estimate, by closing channel switch 575 ~or one frame. As used herein, updating the background noise estimate is defined as partially modifying the old background noise estima~e with a new estimate using, ~or example, a lO~/90~ new-to-lO old estimate ratio. The voice metric values are alsou ed in the channel SNR modlfying proces~ as will subseguently be described.
From the perspective of making a background noise update decision, a ~rame having high energy, which i~
lS typically indicative of a speech frame, could al~o mean that a narrowband noise transient or a sudden increase in tha bac~ground noise level ha occurred. There~ore, the present invention characterizes the frame energy as a voice metric sum, VMSUM, and utiliæes this multi-channel 20 energy parameter to per~orm the updating decision. The procesR utilize~ a voice metric tabla which may be represented as a curve a~ shown in Figure 2.
Figur~ 2 i~ a graph illustrating the characteristic curve o~ the voice metrics for a 25 particular channsl. The horizontal axi~ represents SNR
estimat~ indice~. Each SNR estimate index value represents three-sighth~ (3/8) dB signal-to-noise ratio.
Hence, an S~R estima~e index of lO represent~ 3~,75 dB
S~R. The vertical axis represents voice metric values 30 VM(CC) fnr e~ch of the N channels. Note that a voice metric of 2 i~ produced for an SNR index of 1. Also note that the curva is not linear, since a channel energy has morQ voice-liXe characteristics at higher SNR'~.
First, the raw SNR estimate~ are used to index into the voice metric table to obtain a voice metric 3 ~
r CM00370H
value VM(CC) for each channel. Second, thP individual channel voice metric values are summed to create the to~al of all individual channel voice m~tric values, called the voice metric sum VMSUM. Third, VMSUM i5 05 compared to an UPDATE THRES~OLD represen~ative of a voice metric total that i3 deemed to be noise. If the multi-channel energy parameter VMSUM is less than ~he UPDATE
TH~ESHOLD, the particular frame has very ~ew voice-like characteristicY, and i most probably noise. Therefore, 10 a ba~k~round noise upda~e is performed by closing channel switch 575 for the particular ~rame. The most recent voice metric sum VMSUM is also made available to channel SNR modifier 820 via lin~ 815 for use in the modification algorithm.
In the pre~erred embodiment, the UPDATE THRESHOLD
is set to a total voic~ metric sum value o~ 32. Since the minimum value in the voice metric table is 2, the minimum sum for 14 channel~ is 28. The voice metric table values remain at 2 until an SNR index of 12 (or 20 4.5 dB SNR) is reach~d. This means that an increased level of broadband noisa (individual channel~ each having SNR values not greater than 4.125 dB) will still generate a sum of 28. Sinc~ the UPDATE THRESHOLD o~ 32 would not then be exceeded, t~e broadband noise voice metric will 25 be correctly classified as noise, and a backgxound noise updat~ will be performed. Conversely, any single channel having an SNR index value greater than 24 (or at least 9.O dB SNR) would caus~ the VMSU~ to exceed th~ UPDATE
THRESHOLD, and result ln a voice or narrowband noise 30 burst decision.
Many variations of the voice metric table are possibls, as differen~ type of metrics may be compQnsated for by thQ proper selection o~ the UPDATE
THRESHOLD. Furthermore, the sensitivity of the 35 speech/noise deci~ion may also be cho~en ~or a particular application. For exampla, in the pre~erred embodiment, ..
3~
- r' CM00~70H
the thresAold may be adjusted to accommodate any single channel having an SNR value as sensitive as 4.5 d~ to as insensitive as 15 dB. The corresponding UPDATE THRESHOLD
would then be set wi-thin the range of 29 to 41.
05 In addi~ion to performing the speech/noise decision utilizing voice metrics, voice metric calculator 810 keeps track of the time that has expired since the last background noise update~ An update counter is tested on each frame to see i~ more than a given number 10 o~ frames, each represen~ing a predetermined time, has passed since the previou3 updateO In the preferred embodiment utilizing 10 millisecond frames, if the update counter reaches 100 -corresponding to a timing threshold of 1 second without update~-- an update i~ performed 15 regardless o~ the voice metric decision. However, any timing threshold within the range of 0.5 second to 4 saconds would be practical. As previously mentioned, thi~ timing parameter test i~ used to prevent any sudden, large increases in noise level from being indefinitely 20 interpreted a~ voice~
The basic function of channel SN~ modifier 820 is to eliminate the detrimental effects of narrowband noise bursts on the noise suppression system. A narrowband noise burst may be de~ined as a momentary increase in 25 channel en~rgy for only a few channels. In the pre~erred embodiment, a high energy level.above a 6 dB SNR
thre~hold in fewer than 5 of the upper 10 channels is classified as a narrowband nois2 burst. Such a noise burst would normally create high gain values for only a 30 few number of channels, which results in the "running water" type of background noise flutter described above.
Raw SNR est~mates at 235 are applied to the input of channel SNR modi~ier 820, and modi~ied SNR e~timates are output at 825. Ba~ically, SNR modifier 820 counts 35 the number Or chann~ls which have channel SNR index values which exceed an in~ex t~reshold. In ~he pxe~erred 33~
_ CM00370H
embodiment, the index threshold is set to correspond to an SNR value within the range of 4 dB to 10 dB, preferably 6 dB SNR. If the number of channels is below a predetermined count threshold, then the decision to 05 modify the SNR's is made. The count threshold represents a r~latively few number of channels, i.e., not greater than 40% of the total number o~ channels N. In the pre~exred embodiment, the count threshold is set to 5 of the 10 measured channels. ~uring tha modification 10 process itself, channPl SNR modifier 820 either reduces the SNR of only thosa particular channels having an SNR
index less than a SETBACK THRESHOLD (indicative of a narrowband noise channel), or reduces the SNR of all ~he channels if the voice metric sum i~ less than a metric 15 threshold (indicative of a very weak energy frame).
Hence, the channals containing ~he narrowband noise burst are attenuated so as to prevent them from detrimentally af~ec~ing the gain table look-up function.
SNR threshold block 830 provide a predetermined 20 SNR threshold for each channel which mu~t be exceeded by the modi~ied channel SNR estimates before a high gain value c~n be produced. Only SNR estimates which have a value abovo the SNR threshold are directly applied to the gain tabla sets. Therefore, small background noise 25 fluctuations are no~ allowed to produce gain values which represent voice. This implementation of an SNR threshold essentially present~ an offset in ~he gain risa for channels having low signal-to-noise ratio. Preferably, the SNR threshold would be set within the range of 1.5 dB
30 to 5 dB SNR to eliminat~ minor noise fluctua~ions. The SNR thxeshold may be implemented as a separate element as shown in Figure 1, or i~ may be implemented as a "dead zone" in the characteristic gain curve for each gain table set 590.
FigurQ 3 gxaphically illus~rate~ the function of SNR threshold block 830, as w~ll as tha attenuation ~,?,~ i2 r ~ cM00370H
function o~ the channel gain values in each gain table set. On the horizontal axis, modi~ied SNR estimates are shown in dB as would be output from channel SNR modifier 820 at 825. The vertical axis represents the channel 05 gain ~attenuation) as would be observed at the output of channel gain modifier 250 at 255. A maximum amount of bacXground noise attenuation is achieved for channels having a minimum gain value. Note that SNR threshold block 830 is shown as a "dead zone" or offset in the gain 10 risQ curve o~ approximately 2.25 dB. ~ence, an SNR
estimata mU8t exceed this threshold be~ore the channel gain can rise above the minimum gain level shown. Also note that two curves are illustrated, each having a different minimum gain level~ Upper curve labeled 15 group A represents a 1QW channel group, e.g., consisting of channels 1-4 in the preferred embodiment, while group B represents the higher frequency channels 5=14.
As evident from the graph, the low frequency channels have a minimum gain value of -13.1 dB, while the 20 upper frequency channels have a minimum gain value o~
-20~7 dB. It has been found that less voice quality degradation occurs wh~n the channals are divided into such groups. Although only two different gain curves are used in the preferred embodiment for gain table set 25 number 1, it may prove advantageous to provide each channel with a different characteristic gain curve.
Furthermore, as explained in the referenced Borth patent, multiple gain table sets are used to allow a wider choice o~ channel gain values depending on the particular 30 background noise environment. Noise level quantizer 555 utilize~ hysteresis to select a particular gain table set basQd upon the overall background noi~e estimates. The gain table selection signal, output from noise level quantizer 555, is applied to gain table swi~ch 595 to 35 implement the gain table selection process. Accordingly, one o~ a plurality of gain table sets 5g0 may be chosen .
' CM00370H
as a function of overall average bacXground noise level.
These noise suppression improvements eliminate the variability of the background noise suppression without requiring a large amount of gain smoothing.
05 Background noise attenuation within the range of 10 dB to 25 dB is readily achieved with the present inv~ntion.
With the improvements, the system requires gain smoothing having a time constant of only 10 to 20 milliseconds to obtain a flat or ~whltQI~ residual noise background.
10 Previous technigues required 40 to 60 millisecond time cons~ant gain smoothing, which not only resulted in imperfect flutter reduction, but also substantially degraded the voice quality.
Since the overall operation of the improved noise 15 suppression system is similar to that described in the previous Borth patent, tha generalized flow diagram illustrated in Figure~ 6a/b o~ that patent will be used to describe the pre~ent invention. The general organization of the operation of the present invention 20 may still be organized in three functional groups: noise suppression loop -- seqUQnCe block ~04 oP Figure 6a, which is described in detail in Figure 7a of the Borth patent; automatic gain selector - sequence 615 of Figure 6b, which has been modified for the present invention:
25 and automatic backgroun~ noise estimator -- sequence 621 o~ Figure 6b, which ha~ also been modified in the present invention. The detailed flowcharts of Figure 4a through 4f of the present application may be substituted for sequence blocks 615 and 621 o~ Figure 6b to describe the 30 opQration of improved noise suppression system 800~
Hence, Flgura 6a and 7a of the Borth patent (4,628,529) descri~es the noise suppression loop performed on a samplQ-by-sample basis, while Figures 4a through 4f of the present invention describe the channel gain selection 35 process and the background nois~ estimate update process performed on a ~rame by-frame basis.
r r CM00370H
Referring now to Figure 4a, the operation of improvsd noise suppression system 800 begins from the "YES" output of decision step 614 of the aforementioned Figure 6a. Hence, the actual spectral gain modification 05 function for the particular frame has alr~ady been performed on a sample-by-sample basis utilizing gain values ~rom the previous frameO Sequence 850 serves to generate the SNR estimates available at 235. First o~
all, the channel count CC is set egual to 1 in step 851.
10 Next, the voice metric su~ variable VMSUM is ini~ialized to zero in step 852. Step 853 calculates the raw signal~
to-noise ratio SNR ~or the particular channel as an SNR
estimate index value INDEX(CC). The SNR calculation is simply a division of the per-channel energy estimates 15 (signal-plus-noise) available at 225, by the per-channel background noise estimates (noise) at 325. However, other estimates of the signal-to-noise threshold may alternatively be used. Therefore, step 853 simply divi~es the ~urrent stored channel energy estimate 20 (obtained ~rom ~lowchart step 707 of the aforementioned Figure 7a) by the curr~nt background noise estimate BNE~CC) from the previous ~r~me.
In ~equ~nce 860, the voice metrics are calculated. Firs~, the voice metric table for the 25 particular channel is indexed in step 861 using the raw SMR estimate index INDEX(CC). The voice metric table is read in ~tep ~62 to obtain a voice metric value VM(CC) for the particular channel. This individual channel voice metric value i~ added to ths voice metric sum VMSUM
30 in step 8S3. The channel count CC is incremented in step 864j and tested in step 865. I~ the voice metrics for all N channels have not been calculated, control returns to step 853.
Ssquence 870 illustrates thQ background noise 35 estimata updats deci~ion process performed by voice metric calculator 8I0. The voice metric sum VMSUM i9 ~ CM00370H
compared to UPOATE THRESHOLD in step 871. If VMSU~ is less than or equal to UPDATE THRESHOLD, then tha ~rame is probably a noise frame. TIMER FLAG i5 reset in step 872, and the update counter UC is reset in step 873. Control 05 proceeds to step 878 where the UPDATE FLAG is set true, which means that a background nois~ estimate update will be per~ormed for the current frame.
If VMSUM is greater than the UPDATE THRES~OLD, the ~rame is probably a voice frame. Nevertheless, step 10 874 tests the TIMER FL~G to see if a sudden, loud increase in background noi~e has been interpreted as speech. If the TIMER FLAG is true, the one second time interval was exceeded a number of frames ago, and background noise estimata updating is still required.
15 This is due to the fact that only a partial background noise update is per~ormed for each ~rame. If the TIMER
FLAG is not true, the update counter UC i~ incremented in step 875, and tested in step 876~ If 100 frame~ have occurred since ~he las~ background noise estima~e update, 20the TIMER FLAG is set true in step 877, and the BNE
UPDATE FLAG iq set true in step 878, A series of partial bacXground noise estimats updates are then per~ormed until the voice metric sum VMSUM again falls below the UPDATE THRESHOLD. Note that tha only place in the 2sflowchart that tha TIMER FLAG is reset is in step 872, when the voice metric sum V~SUM again resembles noise.
I~ th~ update counter UC has not reached 100 frames, the instant frame i~ deemed to b~ a voice fram~, and no background noise upda~e is performed.
Re~erring now to sequence 880 o~ Figurec 4b and 4~, the decision to modify ~he channel signal to-noise ratios iq perfor~ed next. An index counter variable I~
i9 initialized in step 881. The channel counter CC is sat equal to 5 in step 882, so a~ to count only the upper 3510 of tho 14 channels having a high energy. The raw SNR
estimate index INDEX(CC) i8 te~ted in step 883 to sea if ~Q~
CMo0370H
it has reached an INDEX THRESHOLD which would correspond to approximately 6 dB SNR. Here, the assumption i~ made that at least 5 of the upper 10 channels of a voice frame should contain ener~y having an SNR of at least 6 dB. If 05the particular channel SNR INDEX(CC) i5 above the INDEX
THRES~OLD, the index count IC is incremented in step 884.
If not, he channel count CC is incremented in step 885 and tested in step 886 to look at the next channel.
When all 10 upper channels have been measured, index count IC represents the number of channels having an SNR estimate index higher than the INDEX THR~SHOLD.
The index count IC is then tested against a COUNT
THRESHOLD in step 887. If IC indicate~ that more channels than the COUNT THRESHOLD, e.g., 5 o~ the upper 15 10 channals, contain surficient energy, ~hen ~he frame is probably a VoicQ ~rame, and tha MODIFY FLAG is set false in step 889 to prevent channel SNR modification. If only a few channals contain high energy, which would be representative of a frame of narrowband noise, then the 20MODIFY FLAG is set true in step 888.
Sequence 890 describes the SNR modification proce~s per~ormed by channel SNR modifier bloc~ 820.
Initially, th~ MODIFY FLAG is te ted in step 891. If it is false, th~ channel SN~ modification process is by-25passed. If the MODIFY FLAG i~ trua, the channel counterCC is initialized in step 892. Nex~, each channel SNR
e~timate index is tested in step 893 to see if it is less than or egual to a SETBACK THRESHOLD. The SETBACK
THRESHOLD, which may have a valuQ corresponding to 6 dB
30SNR, repre~ent~ the maximum SNR estimate which is representative o~ background noise flutter. Only channels having low S~R estimate index pas~ this test.
However, even ir the channel index is greater than the SET~ACK THRES~OLD, the voice metric sum VMSUM is again 35tested in step 894. If VMSUM i~ less than or equal to a k~TRIC THRESHOLD, which corre~ponds to a rapresentativQ
- 3L3~
r r total voice metric o~ a narrowband noise frame, the INDEX(CC) is modified in step 895 ~y setting it equal to the minimum index value of l. The channel counter CC is incremented in step 896 and tes~ed in step 897 ~o s~e i~
05 all the channel have been tested. I~ not, control returns to step ~93 to test the next channel index.
~ence, a frame containing eithPr channel energy fluctuations or narrowband noise is modified ~uch that the frame doe~ not produc~ undesirable gain variations.
Sequance 900 performs the ~unction o~ SNR
threshold block 830. The channel counter CC is initialized in ~tep 901. The SNR index for the particular channel i8 tested against an SNR THRESHOLD in step 902. In the preferred embodiment, the SN~ THRESHOLD
15 represents an index value corresponding ~o 2.25 dB SNR.
If INDEX(CC) is above tha SNR THRESHOLD, it may be used to index the gain table. If not, the index Yalu~ is again set equal to 1 in step 903, which represents the minimum index value. The channel counter CC is 20 incre~ented in step 904 and tested in step 905. This SNR
threshold testing process serve3 to reduce minor background noise variations in all the channels.
Referring now to sequence 910 of Figure 4d, the gain table sets are chosen by noise level quantizer 555 25 and gain table switch 595. In Atep 911, the channel counter CC is initialized, and in step 912, a variable called background noise estimate sum, BNESUM, is initialized. In step 913, the current background noise estimate BNE(CC) i5 obtained ~or each channel, and added 30 to BNESUN in step 914. 5tep 915 increment~ th~ channel counter C~, and step 916 tests the channel counter to see if tha background noi~e estimates for all N channels have bQQn totaled.
In step 917, BNESUM is compared to a ~irst 35 background noise estimat2 threshold. If it is greater than ~NE THRESHOLD 1, then gain tabl~ set number 1 is ' ' , .. .
CMo0370H
selected in step 918. Similarly, step 919 again tes~s BNESUM to see if it is greater than the lower value o~
BNE THRESHOLD 2. If ~NESUM is greater than BNE THRESHOLD
2 but less than BNE THRESHOLD 1, then gain table set 05 number 2 is selected in step 920. Otherwise, gain table set number 3 i~ selected in step 921. Hence, gain table sets 590 are selected as a function of overall average background noise level.
Sequence 930 describes the steps for obtaining lo raw gain values RG(ec) from the gain table sats 590.
Step 931 sets the channel counter CC equal to 1. The selected gain table is indexed in step 932 using the channel SNR estimate index INDEX(CC) whlch has passed the SNR modification and threshold ~ests. The raw gain value 15 RG(CC) is obtained ~rom the selected gain table in step 933, and is then stored in step 934 for use as the gain values for the next frame o~ noise suppression. The channel counter CC is incremen~ed in step 935, and tested in step 936 as before. A3 described in U.5. Patent No.
20 4,630,305, the raw gain values for each channel at 535 are then applied to gain smoothing filter 530 for smoothing on a per-sample bzsiQ.
Finally, sequence 940 describes the actual bacXground noise estimate updating process per~ormed in 25 block 420 o~ Figure 1. Step 941 initially tests the UPDATE FLAG to see if a background noise estimate should be performed. If the UPDATE FLAG is false, then the frame is a voice frame and no background noise upda~e can occur. Otharwise, the background noise update is 30 performed -- which is simulated by closing channel switch 575 -- during a noise frame. In step 942, the UPDATE
FLA~ is reset to ~alse.
Steps 942 through 945 serve to update the current background noise estimate in each of the N channels via 35 the aquation:
$~
~ CMo0~70H
E(i,k) = E(i,k~ SF~(PE(i)-E(i,k-l)], i=1,2, . . . , N
where E(i,k) is the current energy noise estimate for channel (i) at time (k), E(i, k-l) is the old energy 05 noise estimate for channel (i) ak time (k-l), PE(i) is the current pre-processed energy estimate for channel (i), and SF i~ the smoothing factor time constant used in smoothing the background noise estimates. Therefore, E(i, k-l) is stored in energy e~timate storage register 10 585, and the SF term performs the function of smootAing filter 580. In the present embodiment, SF is selected to be 0.1 for a 10 millisecond ~rame duration.
Step 943 initializes the channel count CC to 1.
Step 944 performs the above equation in terms of the 15 current background noise estimate available at 325, ~he old background noise eetimate OL~ BNE(CC) ~tored in energy estimate storage register 585, and the new background noise estimate NEW BNE~CC) available from switch 575. Step 945 increments the channel counter CC, 20 and step 946 tests to see if all N channels have been processed. If true, the background noise estimate update is completed, and operation is returned to step 62~ of Figur~ 6b o~ the a~orementioned Borth patent to reset the sampla counter and incremsnt the ~rame counter. Control 25 then returns to perform noise suppression on a sample-by-sample basis for the next frame~
In review, it can now be seen that the present invention provides the following improvement5: (a) a reduction in background noise ~lutter by offset~ing the 30 gain rise o~ the gain tables until a cextain SNR value is obtained; (b) immunity to narrowband noise bursts through modification Or the SNR estimates based on the volce metrio calculation and the channel energies: and (c) more accurate bacXground noise estimates via performing the 35 update decision based on the overall voice metric and the time interval since the last update.
r CM00370H
While specific embodiments o~ the present invention have been shown and described herein, further modifications and improvement~ may be made by those skilled in the art. Fox exampla, ths operational flow is 05 described herein as performed in real time. However, due to inherent hardware limitations, previous background noise estimates ~or channel gain values may be stored for use in the next frame. All such modification which retain the basic underlying principles disclosed and 0 claims herein are within the scope of this invention.
What is claimed i5:
The noise suppression technique described in the aforementioned patents is the spectral subtraction -- or spectral gain modification -- technigue. Using this ~5 approach, the audio input signal is divided into individual spectral bands by a bank of bandpass filters, and particular spectral bands are attenuated accordingto their noise energy content. A spectral subtraction noise suppression prefilter utilizes an estimate of the 13~;2 " CM00370H
background noise power spectral density ko generate a signal-to-noise ratio (SNR) of the speech in each channel, which, in turn, is used to compute a gain factor for each individual channel. The gain factor is used as 05 a pointer for a look-up table to determine th~
attenuation for that particular ~pectral band. The channels are then attenuated and recombined to produce the noise-suppressed output waveform.
In speGialized applications involving rPlatively 10 high background noise environments, most noise suppression techniques exhibit signi~icant performance limitations. One exampla of uch an application is the vehicle speakerphone option to a cellular mobile radio telephone system, which provides hands-free operation ~or 15 the automobilQ driver. The mobile hands free microphona is typically located at a greater distance from the user, such a~ being mounted overhead on the visor. The more distant microphone delivers a much poorer signal-to-noise ratio to the land-end party due to road and wind noise 20 conditions~ Although the received speech at the land-end is usually intalligible, continuous exposure to such background noise leveIs often increases listener ~atigue.
Although ~ost prior art techniques perform sufficiently well under nominal background noise 25 conditions, the per~ormanc~ of known techniqu2~ becomes severely limited in such specializ~d applications of unusually high background noise. Typical spectral subtraction noise suppre~sion sy tems may reduce the background noise level over the voice ~requency spectrum 30 by as much as 10 dB without seriou~ly a~fecting the speech quality. However, when the prior art tPchniques are used in relatively high background noise environments requiring noise suppression level~ approachtng 20 d~, there is a substantial degradation in the quality 35 characteristics of the voice. Furthermcre, in rapidly-changing high noise environmant~, a sevQre low ~requency r r CM0037OH
noise flutter develops in the output speech siynal which resembles a distant "jet engine roar" sound. This noise flutter is inherent in a spectral subtraction noise suppression system, since the individual channel gain 05 parameters axe continuously being updated in response to the changing background noise environment.
The bacXground noise flutter problem was indirectly addressed but not eliminated through the use of gain smoothing. For example, R.J. McAulay and ~.L.
10 Malpas~, in tha article entitled "Speech Enhancement Using a Soft-Decision Noise SuppresRion Filter", IEEE
Tran~O Acoust., Speech, Siqnal Proces ing, Vol. ASSP-28, No. 2 (April 1980), pp. 137-145, propose the use of gain smoothing on a per-frama basis to avoid the introduction 15 of discontinuitie~ in the output waveform. Since the introduction of gain smoothing can cause the noise suppre~ion prefilter to be slow to respond to a leading edge tran ition (which would result in speech distortion), a weighting ~ac~or of 1 or 1/2 was chosen 20 such that the prefilter responds immediately to an increas~ in gain while tending to smooth any decrease in gain. Unfortunately, excessive gain smoothing still produce~ noticeable detrimental effect~ in voice quality, the primary effect being the apparent introduction of a 25 tail-end ~cho or "noi~e pump" to spoken words. There is also a significan~ reduction in voice amplitude with large amoun~ of gain smoothing.
The noise flutter performance was further improved by th~ technigue of smoothing the noise 30 suppression gain fac~ors for each individual channel on a per~sample basis instead of on a per-frame basis. Per-sampl~ ~moothing, a~ well a~ utilizing different smoothing coef~icient~ for each channel, is described in U.S. Patent No. 4,630/305j entitled "Automatic Gain 35 Selector for a Noise Suppres~ion System." However, none Or the known prior art techniques appreciate that the ~ 3 --rr CM00370H
primary source of the channel gain discontinuities is the inherent fluctuation of background noise in each channel from one frame to the next. In known spectral subtraction syst~ms, even a 2 dB SNR variation would 05 create a few dB o~ gain variation, which is then heard as an annoying background noiss flutter. Hence, the flutter problem has never been effectively sol~ed.
Moreover, narrowband noise -- that which has a high power spectral density in only a ~ew channel~ --10 further complicates the background noise ~lutter problem.~i~ce these ~ew hlgh energy noi~e channels would not be attenuated by the background noise suppressor, the resultan~ audio output has a "running water" type of charac~eristic. Narrowband noise bursts also degrade the 15 accuracy of the background noise update decision required to perform noise ~uppression in changing background noise environment Since the gain factors are chosen by SNR
estimates, which are determined by the speech energy in 20 each channel (signal) and the current background noise energy estimate in each channel (noise), the performance of the entir~ nois~ suppression system i~ based upon the accuracy of the background noise estimate. The statistic~ o~ the background noise are estimated during 25 the time when only background noise is present, such as during the pauses in human speech. There~ore, an accurate speech/noise clas i~ication must be made to determine when such pause~ in speech are occurring.
It is widely known that the energy histogram 30 technigue for distinguishing between background noise and ~paech porform suf~iciently well in normal ambient noise environment~. See, Q.g., W.J. Hes~, "A Pitch Synchronous Digital Feature Extraction System for Phonemic Recognition of Speech," IEEE Trans. Ac u~ Spe_ h, 35 Siqn~l Processin~, Vol. ASSP-24, No. i (February 1976), pp. 14-25. Energy hi~togram~ o~ acou8tic ~ignal8 exhibit r' CMo0370H
a bimodal distribution in which the two modes correspond to noise and speech. Thus, an appropriate threshold can be set between the two modes to provide the speech/noise classi~ication. However, the dis~inction between 05 background noise enargy and unvoiced speech snergy in relatively high background noise environments is wlclear.
Consequently, the task of accurately finding the two modes o~ the energy histogram, and setting the appropriate threshold between them, is extremely 10 dif~icult.
To accommodate changing noi~s backgrounds, McAulay and Malpass implement an adaptive threshold by constantly monitoring the his~ogram energy on a frame-by frama basis, and updating the threshold utilizing 15 different decay factors. Alternatively, U.S. Patent No.
4,630,304 utilizes an energy valley detector to perform the speech/noise decision based upon the post-processed signal energy -- signal en~rgy available at the output of the noise suppression system ~- to determine ~he detected 20 speech minimum. Thus, the accuracy o~ the background noise estimate is improved since it is based upon a much cleansr spe~ch signal.
Howev~r, n~ith~r prior art technique is properly responsive to a sudden, strong increase in background 25 noisQ level. ThesQ background noise estimate updating decision processes interpret a sudden, loud noise level rise as speech, such that no updates are performed. The en~rgy histogram or valley detector has a slow adaptation characteristic which will evantually adapt to the higher 30 noise level. However, this adaptation characteristic does lead to incorrect nois~ updates on the weaker energy portions of speech. Thi~ erroneous decision significantly degrades the performance of t~e noise suppres~ion system.
A ne~d, therefore exists for an improved acoustic noisQ suppression system which addr~sses the problem~ o~
B~
backgrolmd noise fluctuation, narrowband noise bursts, and sudden background noise increases.
Summsly of the Invention Accordingly, it is an object s)f the present invention to provide an improved method and apparatus for suppressing background noise in high background noise environments without significantly degrading the voice quality.
Another object of the present invention is to provide an improved noise suppression system that addresses the background noise fluctuation problemwithout requiring large amounts of gain smoothing.
A further object of the present invention is to provide a spectral 15 subtraction noise suppression system which compensates for the detrimental effects of narrowband noise bursts.
Another object of the present invention is to provide a background noise estimation mechanism which is not misled by low energy portions of speech,20 yet still provides correction for sudden, strong increases in background noise levels.
These and other objects are achieved by the present invention which, briefly described, is an improved noise suppression system for attenuating 25 the background noise from a noisy input signal to produce a ns)ise-suppressedoutput signal by spectral gain modification. The noise suppression system 800 includes a mechanism 210 for separating the input signal into a plurality of pre-processed signals representative of selected frequen~y channels, a mechanism 310for generating an estimate of the signal-to-noise ratio (SNR) in each individual30 channel; a mechanism S90 for producing a gain value for each individual channel by automatically selecting one of a plurality of gain values from a particular gain table in i ~
.
response to the channel SNR estimates, and a mechanism 250 for modi~ing the gain of each of the pluralit~y of pre-processed signals in response to ~he selected gain values to provide a plurality of post-processed noise-suppressed output 5 signals. The improvements of the present invention relate to the addition of an SNR threshold mechanism 830 to eliminate minor gain fluctuations for low SNR
conditions, a voice metric calculator 810 to produce a more accurate background noise estimate update decision, and a channel SNR modifier 820 to suppress narrowband noise bursts.
More specifically, the first aspect of the present invention pertains to the addition of an SNR threshold mechanism 830 for providing a predetermined SNR threshold which the channel SNR estimates must exceed before a gain value above a predefined minimum gain value can be produced.
I~ In the preferred embodiment, the SNR threshold is set at 2.25 dB SNR, such that minor background noise fluctuations do not create step discontinuities in the noise suppression gams.
According to the second aspect of the present invention, a voice 20 metric calculator 810 is utilized to perform the speech/noise classification for the background noise update decision using a two-step process. First, the raw SNR
estimates are used to index a voice metric table to obtain voice metric values for each channel. A voice metric is a measurement of the overall voice-like characteristics of the channel energy. The individual channel voice metric values are summed to create a first multi-channel energy parameter, and then compared to a background noise update threshold. If the voice metric sum does not meet the thresllold, the input frame is deemed to be noise, and a background noise update is performed. Secondly, the time since the occurrence of the previous background estimate update is constantly monitored. If too much time has passed 30 since the last update, e.g., 1 second, then it is assumed that a substantial increase in noise has occurred, and a background noise update is per~ormed regardless of 5 whether it looks like a voice frame. This second test is based on the assumption that speech seldom contains continuous high energy levels in all channels for more than one second, which would be the case for a sudden, loud noise level increase. The voice metric algorithm incorporating the two-step decision processprovides a very accurate background noise estimate update signal.
In the third aspect of the present invention, a channel SNR
modifying mechanism 820 provides a second multi-channel energy parameter in response to the number of upper-charmel SNR estimates which exceed a predetermined energy threshold, e.g., 6 dB SNR. If only a few channels have an 15 energy level above this energy threshold (such as would be the case for a narrowband noise burst), the measured SNR for those particular channels would be reduced. Moreover, if the aforementioned voice metlic sum is less than a metric threshold (which would indicate that the frame was noise), all channels are similarly reduced. This SNR modifying technique is based on the assumption that 20 typical speech exhibits a majority of channels having signal-to-noise ratios of 6 dB
or greater.
Brief Description of the Drauings The features of the present invention which are believed to be novel are set forth with particularity in the appended claims. The invention itself, however, together with further objects and advantages thereof? may best be understood by reference to the following description when taken in conjunction with the accompanying drawings, in which:
Figure 1 is a detailed block diagram illustrating the preferred embodiment of the improved noise .
suppression system according to the present inYent;on;
Figure 2 is a graph representing voice me~ric values output as a func~ion S of SNR estimate index values input for the voice metric calculator block of Figure l;
Figare 3 is a representative gain table graph illustrating the overa~l channel attenuation for particular groups of channels as a function of the SNE~ estima$e;
and Figures 4a through 4f are flowcharts iliustrating the specific sequence of operations performed in accordance with the practice of the preferred embodiment of the present invention.
Detailed Description of the Preferred Ernbodiment Figure 1 is a detailed block diagram of the preferred embodiment of the present invention. ~1 the elements of Figure 1 having re~erence numerals less than 600 correspond to those of U.S. Patent No. 4~628,529 - Borth et al. Refer to the Borth patent for their descriptioll. Ihe additional circuit components having referense numerals greater than 600 represent the improvements to the 20 system, and will be described herein.
Improved noise suppression 800 incorporates changes to the aforementioned Borth noise suppression system in three basic areas: (a) the updating of background noise estimates by voice metric calculator 810; (b) the modification of SNR estimates by channel SNR modifier 820; and (c) utilization 25 of SNR threshold block 830 to offset the gain rise of each channel Each of these improvements will be described in terms of the block diagram of Figure 1, and in terms of the flowchart of Figure 4a - 4f.
Voice metric calculator 810 replaces the valley detector circuitry of the previous system. A voice metric is essentially a measurement of the overall voice-like characteristics o~ the channel ~nergy. Inkhe preferred e~bodiment, voice metric calculator 810 is implemented as a look up table which translates the individual channel SNR estimate~ at 235 into voice metric values. The VoiCQ metric values are used internally to S determins when to update the bacXground noise estimate, by closing channel switch 575 ~or one frame. As used herein, updating the background noise estimate is defined as partially modifying the old background noise estima~e with a new estimate using, ~or example, a lO~/90~ new-to-lO old estimate ratio. The voice metric values are alsou ed in the channel SNR modlfying proces~ as will subseguently be described.
From the perspective of making a background noise update decision, a ~rame having high energy, which i~
lS typically indicative of a speech frame, could al~o mean that a narrowband noise transient or a sudden increase in tha bac~ground noise level ha occurred. There~ore, the present invention characterizes the frame energy as a voice metric sum, VMSUM, and utiliæes this multi-channel 20 energy parameter to per~orm the updating decision. The procesR utilize~ a voice metric tabla which may be represented as a curve a~ shown in Figure 2.
Figur~ 2 i~ a graph illustrating the characteristic curve o~ the voice metrics for a 25 particular channsl. The horizontal axi~ represents SNR
estimat~ indice~. Each SNR estimate index value represents three-sighth~ (3/8) dB signal-to-noise ratio.
Hence, an S~R estima~e index of lO represent~ 3~,75 dB
S~R. The vertical axis represents voice metric values 30 VM(CC) fnr e~ch of the N channels. Note that a voice metric of 2 i~ produced for an SNR index of 1. Also note that the curva is not linear, since a channel energy has morQ voice-liXe characteristics at higher SNR'~.
First, the raw SNR estimate~ are used to index into the voice metric table to obtain a voice metric 3 ~
r CM00370H
value VM(CC) for each channel. Second, thP individual channel voice metric values are summed to create the to~al of all individual channel voice m~tric values, called the voice metric sum VMSUM. Third, VMSUM i5 05 compared to an UPDATE THRES~OLD represen~ative of a voice metric total that i3 deemed to be noise. If the multi-channel energy parameter VMSUM is less than ~he UPDATE
TH~ESHOLD, the particular frame has very ~ew voice-like characteristicY, and i most probably noise. Therefore, 10 a ba~k~round noise upda~e is performed by closing channel switch 575 for the particular ~rame. The most recent voice metric sum VMSUM is also made available to channel SNR modifier 820 via lin~ 815 for use in the modification algorithm.
In the pre~erred embodiment, the UPDATE THRESHOLD
is set to a total voic~ metric sum value o~ 32. Since the minimum value in the voice metric table is 2, the minimum sum for 14 channel~ is 28. The voice metric table values remain at 2 until an SNR index of 12 (or 20 4.5 dB SNR) is reach~d. This means that an increased level of broadband noisa (individual channel~ each having SNR values not greater than 4.125 dB) will still generate a sum of 28. Sinc~ the UPDATE THRESHOLD o~ 32 would not then be exceeded, t~e broadband noise voice metric will 25 be correctly classified as noise, and a backgxound noise updat~ will be performed. Conversely, any single channel having an SNR index value greater than 24 (or at least 9.O dB SNR) would caus~ the VMSU~ to exceed th~ UPDATE
THRESHOLD, and result ln a voice or narrowband noise 30 burst decision.
Many variations of the voice metric table are possibls, as differen~ type of metrics may be compQnsated for by thQ proper selection o~ the UPDATE
THRESHOLD. Furthermore, the sensitivity of the 35 speech/noise deci~ion may also be cho~en ~or a particular application. For exampla, in the pre~erred embodiment, ..
3~
- r' CM00~70H
the thresAold may be adjusted to accommodate any single channel having an SNR value as sensitive as 4.5 d~ to as insensitive as 15 dB. The corresponding UPDATE THRESHOLD
would then be set wi-thin the range of 29 to 41.
05 In addi~ion to performing the speech/noise decision utilizing voice metrics, voice metric calculator 810 keeps track of the time that has expired since the last background noise update~ An update counter is tested on each frame to see i~ more than a given number 10 o~ frames, each represen~ing a predetermined time, has passed since the previou3 updateO In the preferred embodiment utilizing 10 millisecond frames, if the update counter reaches 100 -corresponding to a timing threshold of 1 second without update~-- an update i~ performed 15 regardless o~ the voice metric decision. However, any timing threshold within the range of 0.5 second to 4 saconds would be practical. As previously mentioned, thi~ timing parameter test i~ used to prevent any sudden, large increases in noise level from being indefinitely 20 interpreted a~ voice~
The basic function of channel SN~ modifier 820 is to eliminate the detrimental effects of narrowband noise bursts on the noise suppression system. A narrowband noise burst may be de~ined as a momentary increase in 25 channel en~rgy for only a few channels. In the pre~erred embodiment, a high energy level.above a 6 dB SNR
thre~hold in fewer than 5 of the upper 10 channels is classified as a narrowband nois2 burst. Such a noise burst would normally create high gain values for only a 30 few number of channels, which results in the "running water" type of background noise flutter described above.
Raw SNR est~mates at 235 are applied to the input of channel SNR modi~ier 820, and modi~ied SNR e~timates are output at 825. Ba~ically, SNR modifier 820 counts 35 the number Or chann~ls which have channel SNR index values which exceed an in~ex t~reshold. In ~he pxe~erred 33~
_ CM00370H
embodiment, the index threshold is set to correspond to an SNR value within the range of 4 dB to 10 dB, preferably 6 dB SNR. If the number of channels is below a predetermined count threshold, then the decision to 05 modify the SNR's is made. The count threshold represents a r~latively few number of channels, i.e., not greater than 40% of the total number o~ channels N. In the pre~exred embodiment, the count threshold is set to 5 of the 10 measured channels. ~uring tha modification 10 process itself, channPl SNR modifier 820 either reduces the SNR of only thosa particular channels having an SNR
index less than a SETBACK THRESHOLD (indicative of a narrowband noise channel), or reduces the SNR of all ~he channels if the voice metric sum i~ less than a metric 15 threshold (indicative of a very weak energy frame).
Hence, the channals containing ~he narrowband noise burst are attenuated so as to prevent them from detrimentally af~ec~ing the gain table look-up function.
SNR threshold block 830 provide a predetermined 20 SNR threshold for each channel which mu~t be exceeded by the modi~ied channel SNR estimates before a high gain value c~n be produced. Only SNR estimates which have a value abovo the SNR threshold are directly applied to the gain tabla sets. Therefore, small background noise 25 fluctuations are no~ allowed to produce gain values which represent voice. This implementation of an SNR threshold essentially present~ an offset in ~he gain risa for channels having low signal-to-noise ratio. Preferably, the SNR threshold would be set within the range of 1.5 dB
30 to 5 dB SNR to eliminat~ minor noise fluctua~ions. The SNR thxeshold may be implemented as a separate element as shown in Figure 1, or i~ may be implemented as a "dead zone" in the characteristic gain curve for each gain table set 590.
FigurQ 3 gxaphically illus~rate~ the function of SNR threshold block 830, as w~ll as tha attenuation ~,?,~ i2 r ~ cM00370H
function o~ the channel gain values in each gain table set. On the horizontal axis, modi~ied SNR estimates are shown in dB as would be output from channel SNR modifier 820 at 825. The vertical axis represents the channel 05 gain ~attenuation) as would be observed at the output of channel gain modifier 250 at 255. A maximum amount of bacXground noise attenuation is achieved for channels having a minimum gain value. Note that SNR threshold block 830 is shown as a "dead zone" or offset in the gain 10 risQ curve o~ approximately 2.25 dB. ~ence, an SNR
estimata mU8t exceed this threshold be~ore the channel gain can rise above the minimum gain level shown. Also note that two curves are illustrated, each having a different minimum gain level~ Upper curve labeled 15 group A represents a 1QW channel group, e.g., consisting of channels 1-4 in the preferred embodiment, while group B represents the higher frequency channels 5=14.
As evident from the graph, the low frequency channels have a minimum gain value of -13.1 dB, while the 20 upper frequency channels have a minimum gain value o~
-20~7 dB. It has been found that less voice quality degradation occurs wh~n the channals are divided into such groups. Although only two different gain curves are used in the preferred embodiment for gain table set 25 number 1, it may prove advantageous to provide each channel with a different characteristic gain curve.
Furthermore, as explained in the referenced Borth patent, multiple gain table sets are used to allow a wider choice o~ channel gain values depending on the particular 30 background noise environment. Noise level quantizer 555 utilize~ hysteresis to select a particular gain table set basQd upon the overall background noi~e estimates. The gain table selection signal, output from noise level quantizer 555, is applied to gain table swi~ch 595 to 35 implement the gain table selection process. Accordingly, one o~ a plurality of gain table sets 5g0 may be chosen .
' CM00370H
as a function of overall average bacXground noise level.
These noise suppression improvements eliminate the variability of the background noise suppression without requiring a large amount of gain smoothing.
05 Background noise attenuation within the range of 10 dB to 25 dB is readily achieved with the present inv~ntion.
With the improvements, the system requires gain smoothing having a time constant of only 10 to 20 milliseconds to obtain a flat or ~whltQI~ residual noise background.
10 Previous technigues required 40 to 60 millisecond time cons~ant gain smoothing, which not only resulted in imperfect flutter reduction, but also substantially degraded the voice quality.
Since the overall operation of the improved noise 15 suppression system is similar to that described in the previous Borth patent, tha generalized flow diagram illustrated in Figure~ 6a/b o~ that patent will be used to describe the pre~ent invention. The general organization of the operation of the present invention 20 may still be organized in three functional groups: noise suppression loop -- seqUQnCe block ~04 oP Figure 6a, which is described in detail in Figure 7a of the Borth patent; automatic gain selector - sequence 615 of Figure 6b, which has been modified for the present invention:
25 and automatic backgroun~ noise estimator -- sequence 621 o~ Figure 6b, which ha~ also been modified in the present invention. The detailed flowcharts of Figure 4a through 4f of the present application may be substituted for sequence blocks 615 and 621 o~ Figure 6b to describe the 30 opQration of improved noise suppression system 800~
Hence, Flgura 6a and 7a of the Borth patent (4,628,529) descri~es the noise suppression loop performed on a samplQ-by-sample basis, while Figures 4a through 4f of the present invention describe the channel gain selection 35 process and the background nois~ estimate update process performed on a ~rame by-frame basis.
r r CM00370H
Referring now to Figure 4a, the operation of improvsd noise suppression system 800 begins from the "YES" output of decision step 614 of the aforementioned Figure 6a. Hence, the actual spectral gain modification 05 function for the particular frame has alr~ady been performed on a sample-by-sample basis utilizing gain values ~rom the previous frameO Sequence 850 serves to generate the SNR estimates available at 235. First o~
all, the channel count CC is set egual to 1 in step 851.
10 Next, the voice metric su~ variable VMSUM is ini~ialized to zero in step 852. Step 853 calculates the raw signal~
to-noise ratio SNR ~or the particular channel as an SNR
estimate index value INDEX(CC). The SNR calculation is simply a division of the per-channel energy estimates 15 (signal-plus-noise) available at 225, by the per-channel background noise estimates (noise) at 325. However, other estimates of the signal-to-noise threshold may alternatively be used. Therefore, step 853 simply divi~es the ~urrent stored channel energy estimate 20 (obtained ~rom ~lowchart step 707 of the aforementioned Figure 7a) by the curr~nt background noise estimate BNE~CC) from the previous ~r~me.
In ~equ~nce 860, the voice metrics are calculated. Firs~, the voice metric table for the 25 particular channel is indexed in step 861 using the raw SMR estimate index INDEX(CC). The voice metric table is read in ~tep ~62 to obtain a voice metric value VM(CC) for the particular channel. This individual channel voice metric value i~ added to ths voice metric sum VMSUM
30 in step 8S3. The channel count CC is incremented in step 864j and tested in step 865. I~ the voice metrics for all N channels have not been calculated, control returns to step 853.
Ssquence 870 illustrates thQ background noise 35 estimata updats deci~ion process performed by voice metric calculator 8I0. The voice metric sum VMSUM i9 ~ CM00370H
compared to UPOATE THRESHOLD in step 871. If VMSU~ is less than or equal to UPDATE THRESHOLD, then tha ~rame is probably a noise frame. TIMER FLAG i5 reset in step 872, and the update counter UC is reset in step 873. Control 05 proceeds to step 878 where the UPDATE FLAG is set true, which means that a background nois~ estimate update will be per~ormed for the current frame.
If VMSUM is greater than the UPDATE THRES~OLD, the ~rame is probably a voice frame. Nevertheless, step 10 874 tests the TIMER FL~G to see if a sudden, loud increase in background noi~e has been interpreted as speech. If the TIMER FLAG is true, the one second time interval was exceeded a number of frames ago, and background noise estimata updating is still required.
15 This is due to the fact that only a partial background noise update is per~ormed for each ~rame. If the TIMER
FLAG is not true, the update counter UC i~ incremented in step 875, and tested in step 876~ If 100 frame~ have occurred since ~he las~ background noise estima~e update, 20the TIMER FLAG is set true in step 877, and the BNE
UPDATE FLAG iq set true in step 878, A series of partial bacXground noise estimats updates are then per~ormed until the voice metric sum VMSUM again falls below the UPDATE THRESHOLD. Note that tha only place in the 2sflowchart that tha TIMER FLAG is reset is in step 872, when the voice metric sum V~SUM again resembles noise.
I~ th~ update counter UC has not reached 100 frames, the instant frame i~ deemed to b~ a voice fram~, and no background noise upda~e is performed.
Re~erring now to sequence 880 o~ Figurec 4b and 4~, the decision to modify ~he channel signal to-noise ratios iq perfor~ed next. An index counter variable I~
i9 initialized in step 881. The channel counter CC is sat equal to 5 in step 882, so a~ to count only the upper 3510 of tho 14 channels having a high energy. The raw SNR
estimate index INDEX(CC) i8 te~ted in step 883 to sea if ~Q~
CMo0370H
it has reached an INDEX THRESHOLD which would correspond to approximately 6 dB SNR. Here, the assumption i~ made that at least 5 of the upper 10 channels of a voice frame should contain ener~y having an SNR of at least 6 dB. If 05the particular channel SNR INDEX(CC) i5 above the INDEX
THRES~OLD, the index count IC is incremented in step 884.
If not, he channel count CC is incremented in step 885 and tested in step 886 to look at the next channel.
When all 10 upper channels have been measured, index count IC represents the number of channels having an SNR estimate index higher than the INDEX THR~SHOLD.
The index count IC is then tested against a COUNT
THRESHOLD in step 887. If IC indicate~ that more channels than the COUNT THRESHOLD, e.g., 5 o~ the upper 15 10 channals, contain surficient energy, ~hen ~he frame is probably a VoicQ ~rame, and tha MODIFY FLAG is set false in step 889 to prevent channel SNR modification. If only a few channals contain high energy, which would be representative of a frame of narrowband noise, then the 20MODIFY FLAG is set true in step 888.
Sequence 890 describes the SNR modification proce~s per~ormed by channel SNR modifier bloc~ 820.
Initially, th~ MODIFY FLAG is te ted in step 891. If it is false, th~ channel SN~ modification process is by-25passed. If the MODIFY FLAG i~ trua, the channel counterCC is initialized in step 892. Nex~, each channel SNR
e~timate index is tested in step 893 to see if it is less than or egual to a SETBACK THRESHOLD. The SETBACK
THRESHOLD, which may have a valuQ corresponding to 6 dB
30SNR, repre~ent~ the maximum SNR estimate which is representative o~ background noise flutter. Only channels having low S~R estimate index pas~ this test.
However, even ir the channel index is greater than the SET~ACK THRES~OLD, the voice metric sum VMSUM is again 35tested in step 894. If VMSUM i~ less than or equal to a k~TRIC THRESHOLD, which corre~ponds to a rapresentativQ
- 3L3~
r r total voice metric o~ a narrowband noise frame, the INDEX(CC) is modified in step 895 ~y setting it equal to the minimum index value of l. The channel counter CC is incremented in step 896 and tes~ed in step 897 ~o s~e i~
05 all the channel have been tested. I~ not, control returns to step ~93 to test the next channel index.
~ence, a frame containing eithPr channel energy fluctuations or narrowband noise is modified ~uch that the frame doe~ not produc~ undesirable gain variations.
Sequance 900 performs the ~unction o~ SNR
threshold block 830. The channel counter CC is initialized in ~tep 901. The SNR index for the particular channel i8 tested against an SNR THRESHOLD in step 902. In the preferred embodiment, the SN~ THRESHOLD
15 represents an index value corresponding ~o 2.25 dB SNR.
If INDEX(CC) is above tha SNR THRESHOLD, it may be used to index the gain table. If not, the index Yalu~ is again set equal to 1 in step 903, which represents the minimum index value. The channel counter CC is 20 incre~ented in step 904 and tested in step 905. This SNR
threshold testing process serve3 to reduce minor background noise variations in all the channels.
Referring now to sequence 910 of Figure 4d, the gain table sets are chosen by noise level quantizer 555 25 and gain table switch 595. In Atep 911, the channel counter CC is initialized, and in step 912, a variable called background noise estimate sum, BNESUM, is initialized. In step 913, the current background noise estimate BNE(CC) i5 obtained ~or each channel, and added 30 to BNESUN in step 914. 5tep 915 increment~ th~ channel counter C~, and step 916 tests the channel counter to see if tha background noi~e estimates for all N channels have bQQn totaled.
In step 917, BNESUM is compared to a ~irst 35 background noise estimat2 threshold. If it is greater than ~NE THRESHOLD 1, then gain tabl~ set number 1 is ' ' , .. .
CMo0370H
selected in step 918. Similarly, step 919 again tes~s BNESUM to see if it is greater than the lower value o~
BNE THRESHOLD 2. If ~NESUM is greater than BNE THRESHOLD
2 but less than BNE THRESHOLD 1, then gain table set 05 number 2 is selected in step 920. Otherwise, gain table set number 3 i~ selected in step 921. Hence, gain table sets 590 are selected as a function of overall average background noise level.
Sequence 930 describes the steps for obtaining lo raw gain values RG(ec) from the gain table sats 590.
Step 931 sets the channel counter CC equal to 1. The selected gain table is indexed in step 932 using the channel SNR estimate index INDEX(CC) whlch has passed the SNR modification and threshold ~ests. The raw gain value 15 RG(CC) is obtained ~rom the selected gain table in step 933, and is then stored in step 934 for use as the gain values for the next frame o~ noise suppression. The channel counter CC is incremen~ed in step 935, and tested in step 936 as before. A3 described in U.5. Patent No.
20 4,630,305, the raw gain values for each channel at 535 are then applied to gain smoothing filter 530 for smoothing on a per-sample bzsiQ.
Finally, sequence 940 describes the actual bacXground noise estimate updating process per~ormed in 25 block 420 o~ Figure 1. Step 941 initially tests the UPDATE FLAG to see if a background noise estimate should be performed. If the UPDATE FLAG is false, then the frame is a voice frame and no background noise upda~e can occur. Otharwise, the background noise update is 30 performed -- which is simulated by closing channel switch 575 -- during a noise frame. In step 942, the UPDATE
FLA~ is reset to ~alse.
Steps 942 through 945 serve to update the current background noise estimate in each of the N channels via 35 the aquation:
$~
~ CMo0~70H
E(i,k) = E(i,k~ SF~(PE(i)-E(i,k-l)], i=1,2, . . . , N
where E(i,k) is the current energy noise estimate for channel (i) at time (k), E(i, k-l) is the old energy 05 noise estimate for channel (i) ak time (k-l), PE(i) is the current pre-processed energy estimate for channel (i), and SF i~ the smoothing factor time constant used in smoothing the background noise estimates. Therefore, E(i, k-l) is stored in energy e~timate storage register 10 585, and the SF term performs the function of smootAing filter 580. In the present embodiment, SF is selected to be 0.1 for a 10 millisecond ~rame duration.
Step 943 initializes the channel count CC to 1.
Step 944 performs the above equation in terms of the 15 current background noise estimate available at 325, ~he old background noise eetimate OL~ BNE(CC) ~tored in energy estimate storage register 585, and the new background noise estimate NEW BNE~CC) available from switch 575. Step 945 increments the channel counter CC, 20 and step 946 tests to see if all N channels have been processed. If true, the background noise estimate update is completed, and operation is returned to step 62~ of Figur~ 6b o~ the a~orementioned Borth patent to reset the sampla counter and incremsnt the ~rame counter. Control 25 then returns to perform noise suppression on a sample-by-sample basis for the next frame~
In review, it can now be seen that the present invention provides the following improvement5: (a) a reduction in background noise ~lutter by offset~ing the 30 gain rise o~ the gain tables until a cextain SNR value is obtained; (b) immunity to narrowband noise bursts through modification Or the SNR estimates based on the volce metrio calculation and the channel energies: and (c) more accurate bacXground noise estimates via performing the 35 update decision based on the overall voice metric and the time interval since the last update.
r CM00370H
While specific embodiments o~ the present invention have been shown and described herein, further modifications and improvement~ may be made by those skilled in the art. Fox exampla, ths operational flow is 05 described herein as performed in real time. However, due to inherent hardware limitations, previous background noise estimates ~or channel gain values may be stored for use in the next frame. All such modification which retain the basic underlying principles disclosed and 0 claims herein are within the scope of this invention.
What is claimed i5:
Claims (50)
1. An improved noise suppression system for attenuating the background noise from a noisy input signal to produce a noise suppressed output signal, said noise suppression system comprising:
means for separating the input signal into a plurality of preprocessed signals representative of selected frequency channels;
means for generating estimates of the signal-plus-noise energy and the noise energy in each individual channel;
means for producing a gain value for each individual channel in response to said channel energy estimates, said gain values having a minimum gain value for each channel, said gain value producing means including threshold means for allowing gain values above said minimum gain value to be produced only when said signal-plus-noise energy estimates exceed said noise energy estimates by a predetermined amount; and means for modifying the gain of each of said plurality of preprocessed signals in response to said gain values to provide a plurality of post-processedsignals.
means for separating the input signal into a plurality of preprocessed signals representative of selected frequency channels;
means for generating estimates of the signal-plus-noise energy and the noise energy in each individual channel;
means for producing a gain value for each individual channel in response to said channel energy estimates, said gain values having a minimum gain value for each channel, said gain value producing means including threshold means for allowing gain values above said minimum gain value to be produced only when said signal-plus-noise energy estimates exceed said noise energy estimates by a predetermined amount; and means for modifying the gain of each of said plurality of preprocessed signals in response to said gain values to provide a plurality of post-processedsignals.
2. The noise suppression system according to claim 1, wherein said gain value producing means produces gain values based upon the signal-to-noise ratio (SNR) of said channel energy estimates, and wherein said SNR estimates are compared with a predefined SNR threshold such that channels having SNR estimates below said SNR threshold produce minimum gain values.
3. The noise suppression system according to claim 2, wherein said predefined SNR threshold corresponds to an SNR value within the range of 1.5 dB to 5 dB SNR.
4. The noise suppression system according to claim 3, wherein said predefined SNR threshold corresponds to an SNR value of approximately 2.25 dB SNR.
5. The noise suppression system according to claim 1, wherein said gain modifying means provides a maximum amount of attenuation of the pre-processed signal in a particular channel having a minimum gain value.
6. The noise suppression system according to claim 1, wherein gain values produce a higher amount of attenuation for high frequency channels than low frequency channels.
7. The noise suppression system according to claim 1, wherein said gain value producing means further includes a plurality of gain tables, each gain table having predetermined individual channel gain values corresponding to said individual channel energy estimates, and gain table selection means for automatically selecting one of said plurality of gain tables as a function of the overall average background noise level of said input signal.
8. The noise suppression system according to claim 1, further includes means for combining said plurality of post-processed signals to produce said noise-suppressed output signal.
9. An improved noise suppression system for attenuating the background noise from a noisy input signal to produce a noise-suppressed output signal, said noise suppression system comprising:
means for separating the input signal into a plurality of pre-processed signals representative of selected frequency channels;
means for generating and storing an estimate of the background noise power spectral density of said pre-processed signals, said background noise estimate generating means including means for modifying said background noise estimate in response to a timing parameter indicative of the time interval since the previous background noise estimate modification;
means for generating an estimate of the signal-to-noise ratio (SNR) in each individual channel based upon said modified background noise estimates;
means for producing a gain value for each individual channel in response to said channel SNR estimates; and means for modifying the gain of each of said plurality of pre-processed signals in response to said gain values to provide a plurality of post-processed signals.
means for separating the input signal into a plurality of pre-processed signals representative of selected frequency channels;
means for generating and storing an estimate of the background noise power spectral density of said pre-processed signals, said background noise estimate generating means including means for modifying said background noise estimate in response to a timing parameter indicative of the time interval since the previous background noise estimate modification;
means for generating an estimate of the signal-to-noise ratio (SNR) in each individual channel based upon said modified background noise estimates;
means for producing a gain value for each individual channel in response to said channel SNR estimates; and means for modifying the gain of each of said plurality of pre-processed signals in response to said gain values to provide a plurality of post-processed signals.
10. The noise suppression system according to claim 9, wherein said background noise estimate modifying means includes means for producing said timing parameter, and means for comparing said timing parameter to a predetermined timing threshold such that a background noise estimate modification is performed when said timing parameter exceeds said timing threshold.
11. The noise suppression system according to claim 10, wherein said predetermined timing threshold is in the range of 0.5 second to 4 seconds.
12. The noise suppression system according to claim 11, wherein said predetermined timing threshold is approximately equal to 1 second.
13. The noise suppression system according to claim 10, wherein said background noise estimate modifying means further includes means for generating an estimate of the energy in each individual channel, and means for producing a multi-channel energy parameter in response to the total value of all individual channel energy estimates.
14. The noise suppression system according to claim 13, wherein said background noise estimate modifying means further includes means for comparing said multi-channel energy parameter to a predetermined energy threshold such that a background noise estimate modification is performed when said multi-channel energy parameter is less than said energy threshold.
15. The noise suppression system according to claim 13, wherein said multi-channel energy parameter is generated by translating said individual channel SNR estimates into individual channel voice metrics and summing the individual channel voice metrics, the voice metric sum being a measurement of the overall voice-like characteristics of the energy in all channels.
16. The noise suppression system according to claim 14, wherein said background noise estimate modifying means modifies said background noise estimates in response to said timing parameter regardless of said multi-channel energy parameter.
17. The noise suppression system according to claim 13, wherein said multi-channel energy parameter producing means accommodates for minor variations in individual channel energy estimates such that said minor variations do not significantly affect said multi-channel energy parameter.
18. The noise suppression system according to claim 14, wherein said predetermined energy threshold is set such that a background noise estimate modification is performed if all channels exhibit individual SNR
values less than 6 dB SNR.
values less than 6 dB SNR.
19. The noise suppression system according to claim 14, wherein said predetermined energy threshold is set such that a background noise estimate modification is not performed if any single channel exhibits an SNR
value of at least 6 dB SNR.
value of at least 6 dB SNR.
20. The noise suppression system according to claim 9, wherein said gain value producing means further includes a plurality of gain tables, each gain table having predetermined individual channel gain values corresponding to various individual channel SNR estimates, and gain table selection means for automatically selecting one of said plurality of gain tables as a function of the overall average background noise level of said input signal.
21. The noise suppression system according to claim 9, further includes means for combining said plurality of post-processed signals to produce said noise-suppressed output signal.
22. An improved noise suppression system for attenuating the background noise from a noisy input signal to produce a noise-suppressed output signal, said noise suppression system comprising:
means for separating the input signal into a plurality of pre-processed signals representative of a number N of selected frequency channels means for generating an estimate of the energy in each individual channel;
means for monitoring said channel energy estimates and for distinguishing narrowband noise bursts from speech energy and background noise energy, thereby producing a modification signal;
means for selectively modifying said channel energy estimates in response to said modification signal such that channel energy estimates representative of narrowband noise bursts are modified;
means for producing a gain value for each individual channel in response to each modified channel energy estimate;
and means for modifying the gain of each of said plurality of pre-processed signals in response to said gain values to provide a plurality of post-processed signals.
means for separating the input signal into a plurality of pre-processed signals representative of a number N of selected frequency channels means for generating an estimate of the energy in each individual channel;
means for monitoring said channel energy estimates and for distinguishing narrowband noise bursts from speech energy and background noise energy, thereby producing a modification signal;
means for selectively modifying said channel energy estimates in response to said modification signal such that channel energy estimates representative of narrowband noise bursts are modified;
means for producing a gain value for each individual channel in response to each modified channel energy estimate;
and means for modifying the gain of each of said plurality of pre-processed signals in response to said gain values to provide a plurality of post-processed signals.
23. The noise suppression system according to claim 22, wherein said modification signal is indicative of the total number of individual channels having energy estimates exceeding a predetermined energy threshold.
24. The noise suppression system according to claim 23, wherein said predetermined energy threshold corresponds to a signal-to-noise ratio (SNR) value within the range of 4 dB to 10 dB SNR.
25. The noise suppression system according to claim 24, wherein said predetermined energy threshold corresponds to an SNR value of approximately 6 dB SNR.
26. The noise suppression system according to claim 23, wherein said channel energy estimate modifying means includes means for comparing said modification signal to a predetermined count threshold such that a channel energy estimate modification is performed when said total number of individual channels is less than said count threshold.
27. The noise suppression system according to claim 26, wherein said predetermined count threshold corresponds to less than 40% X N.
28. The noise suppression system according to claim 22, wherein said gain modifying means provides a maximum amount of attenuation of the pre-processed signal in a particular channel having a modified channel energy estimate.
29. The noise suppression system according to claim 22, wherein said gain value producing means further includes a plurality of gain tables, each gain table having predetermined individual channel gain values corresponding to various individual channel energy estimates, and gain table selection means for automatically selecting one of said plurality of gain tablesas a function of the overall average background noise level of said input signal.
30. The noise suppression system according to claim 22, further includes means for combining said plurality of post-processed signals to produce said noise-suppressed output signal.
31. An improved method of attenuating the background noise from a noisy input signal to produce a noise-suppressed output signal in a noise suppression system comprising the steps of:
separating the input signal into a plurality of preprocessed signals representative of a number N of selected frequency channels;
generating an estimate of the energy in each individual channel;
generating and storing an estimate of the background noise power spectral density of said pre-processed signals;
generating an estimate of the signal-to-noise ratio (SNR) in each individual channel based upon said background noise estimates and said channel energy estimates;
producing a gain value for each individual channel in response to said channel SNR estimates, said gain values having a range of minimal values, said gain value producing step including the steps of providing a predefined SNR threshold and comparing said channel SNR estimates to said predefined SNR threshold such that channels having SNR estimates below said SNR
threshold produce gain values within said minimal range; and modifying the gain of each of said plurality of preprocessed signals in response to said gain values to provide a plurality of post-processed signals.
separating the input signal into a plurality of preprocessed signals representative of a number N of selected frequency channels;
generating an estimate of the energy in each individual channel;
generating and storing an estimate of the background noise power spectral density of said pre-processed signals;
generating an estimate of the signal-to-noise ratio (SNR) in each individual channel based upon said background noise estimates and said channel energy estimates;
producing a gain value for each individual channel in response to said channel SNR estimates, said gain values having a range of minimal values, said gain value producing step including the steps of providing a predefined SNR threshold and comparing said channel SNR estimates to said predefined SNR threshold such that channels having SNR estimates below said SNR
threshold produce gain values within said minimal range; and modifying the gain of each of said plurality of preprocessed signals in response to said gain values to provide a plurality of post-processed signals.
32. The method according to claim 31, wherein said predefined SNR threshold corresponds to an SNR value within the range of 1.5 dB to 5 dB SNR.
33. The method according to claim 31, wherein said gain modifying step provides a maximum amount of attenuation of the pre-processed signal in a particular channel having a gain value within said minimal range.
34. The method according to claim 31, including the step of modifying said background noise estimate in response to a timing parameter indicative of the time interval since the previous background noise estimate modification.
35. The method according to claim 34, wherein said background noise estimate modifying step includes the steps of producing said timing parameter and comparing said timing parameter to a predetermined timing threshold such that a background noise estimate modification is performed when said timing parameter exceeds said timing threshold.
36. The method according to claim 35, wherein said predetermined timing threshold is in the range of 0.5 second to 4 seconds.
37. The method according to claim 34, wherein said background noise estimate modifying step further includes the step of producing a multi-channel energy parameter in response to the total value of all individual channel SNR estimates.
38. The method according to claim 37, wherein said background noise estimate modifying step further includes the step of comparing said multi-channel energy parameter to a predetermined energy threshold such that a background noise estimate modification is performed when said multi-channel energy parameter is less than said energy threshold.
39. The method according to claim 38, wherein said multi-channel energy parameter is generated by translating said individual channel SNR
estimates into individual channel voice metrics and summing the individual channel voice metrics, the voice metric sum being a measurement of the overall voice-like characteristics of the energy in all channels.
estimates into individual channel voice metrics and summing the individual channel voice metrics, the voice metric sum being a measurement of the overall voice-like characteristics of the energy in all channels.
40. The method according to claim 38, wherein said background noise estimate modifying step modifies said background noise estimates in response to said timing parameter regardless of said multi-channel energy parameter.
41. The method according to claim 38, wherein said predetermined energy threshold is set such that a background noise estimate modification is performed if all channels exhibit individual SNR values less than 6 dB SNR.
42. The method according to claim 38, wherein said predetermined energy threshold is set such that a background noise estimate modification is not performed if any single channel exhibits an SNR value of at least 6 dB
SNR.
SNR.
43. The method according to claim 31, including the steps of monitoring said channel SNR estimates and distinguishing narrowband noise bursts from speech energy and background noise energy thereby producing a modification signal, and selectively modifying said channel SNR estimates in response to said modification signal such that channel SNR estimates representative of narrowband noise bursts are modified.
44. The method according to claim 43, wherein said modification signal is indicative of the total number of individual channels having SNR
estimates exceeding a predetermined modification threshold.
estimates exceeding a predetermined modification threshold.
45. The method according to claim 44, wherein said predetermined modification threshold corresponds to an SNR value within the range of 4 dB
to 10 dB SNR.
to 10 dB SNR.
46. The method according to claim 44, wherein said channel SNR
estimate modifying step includes the step of comparing said modification signal to a predetermined count threshold such that a channel SNR estimate modification is performed when said total number of individual channels is less than said count threshold.
estimate modifying step includes the step of comparing said modification signal to a predetermined count threshold such that a channel SNR estimate modification is performed when said total number of individual channels is less than said count threshold.
47. The method according to claim 46, wherein said predetermined count threshold corresponds to less than 40% X N.
48. The method according to claim 43, wherein said gain modifying step provides a maximum amount of attenuation of the pre-processed signal in a particular channel having a modified channel SNR estimate.
49. The method according to claim 31, wherein said gain value producing step further includes the step of automatically selecting one of a plurality of gain tables as a function of the overall average background noise level of said input signal, each gain table having predetermined individual channel gain values corresponding to various individual channel SNR
estimates.
estimates.
50. The method according to claim 31, further includes the step of combining said plurality of post-processed signals to produce said noise-suppressed output signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CA000612604A CA1308362C (en) | 1989-09-22 | 1989-09-22 | Noise suppression system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CA000612604A CA1308362C (en) | 1989-09-22 | 1989-09-22 | Noise suppression system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CA1308362C true CA1308362C (en) | 1992-10-06 |
Family
ID=4140656
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CA000612604A Expired - Lifetime CA1308362C (en) | 1989-09-22 | 1989-09-22 | Noise suppression system |
Country Status (1)
| Country | Link |
|---|---|
| CA (1) | CA1308362C (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112185410A (en) * | 2020-10-21 | 2021-01-05 | 北京猿力未来科技有限公司 | Audio processing method and device |
-
1989
- 1989-09-22 CA CA000612604A patent/CA1308362C/en not_active Expired - Lifetime
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112185410A (en) * | 2020-10-21 | 2021-01-05 | 北京猿力未来科技有限公司 | Audio processing method and device |
| CN112185410B (en) * | 2020-10-21 | 2024-04-30 | 北京猿力未来科技有限公司 | Audio processing method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US4811404A (en) | Noise suppression system | |
| US4628529A (en) | Noise suppression system | |
| EP0226613B1 (en) | Noise supression system | |
| US4630305A (en) | Automatic gain selector for a noise suppression system | |
| US5708754A (en) | Method for real-time reduction of voice telecommunications noise not measurable at its source | |
| US5642464A (en) | Methods and apparatus for noise conditioning in digital speech compression systems using linear predictive coding | |
| KR100546468B1 (en) | Noise suppression system and method | |
| US7058572B1 (en) | Reducing acoustic noise in wireless and landline based telephony | |
| US4630304A (en) | Automatic background noise estimator for a noise suppression system | |
| KR100335162B1 (en) | Noise reduction method of noise signal and noise section detection method | |
| KR100851716B1 (en) | Noise suppression based on bark band weiner filtering and modified doblinger noise estimate | |
| US5839101A (en) | Noise suppressor and method for suppressing background noise in noisy speech, and a mobile station | |
| US5544250A (en) | Noise suppression system and method therefor | |
| US7454010B1 (en) | Noise reduction and comfort noise gain control using bark band weiner filter and linear attenuation | |
| KR100909679B1 (en) | Enhanced Artificial Bandwidth Expansion System and Method | |
| EP1080463B1 (en) | Signal noise reduction by spectral subtraction using spectrum dependent exponential gain function averaging | |
| KR20010033994A (en) | Methods and apparatus for providing comfort noise in communications systems | |
| JP2003500936A (en) | Improving near-end audio signals in echo suppression systems | |
| EP1279163A1 (en) | Speech presence measurement detection techniques | |
| CA2247429C (en) | Apparatus and method for non-linear processing in a communication system | |
| CA1308362C (en) | Noise suppression system | |
| EP0588526B1 (en) | A method of and system for noise suppression |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| MKEX | Expiry |