Ex Parte Kallio et al

Patent Trials and Appeals Board

Apr 11, 2013

10341332 - (D) (P.T.A.B. Apr. 11, 2013)

UNITED STATES PATENT AND TRADEMARKOFFICE
UNITED STATES DEPARTMENT OF COMMERCE
United States Patent and Trademark Office
Address: COMMISSIONER FOR PATENTS
P.O. Box 1450
Alexandria, Virginia 22313-1450
www.uspto.gov
APPLICATION NO. FILING DATE FIRST NAMED INVENTOR ATTORNEY DOCKET NO. CONFIRMATION NO.
10/341,332 01/10/2003 Loura Kallio 944-001.096 7106
4955 7590 04/12/2013
WARE, FRESSOLA, MAGUIRE & BARBER LLP
BRADFORD GREEN, BUILDING 5
755 MAIN STREET, P O BOX 224
MONROE, CT 06468
EXAMINER
HAN, QI
ART UNIT PAPER NUMBER
2659
MAIL DATE DELIVERY MODE
04/12/2013 PAPER
Please find below and/or attached an Office communication concerning this application or proceeding.
The time period for reply, if any, is set in the attached communication.
PTOL-90A (Rev. 04/07)

UNITED STATES PATENT AND TRADEMARK OFFICE
________________

BEFORE THE PATENT TRIAL AND APPEAL BOARD
________________

Ex parte LOURA KALLIO, PAAVO ALKU,
KIMMO KAYHKO, MATTI KAJALA, and
PAIVI VALVE

________________

Appeal 2010-006945
Application 10/341,332
Technology Center 2600
________________

Before JEAN R. HOMERE, MICHAEL J. STRAUSS, and JOHN G. NEW,
Administrative Patent Judges.

NEW, Administrative Patent Judge.

DECISION ON APPEAL

Appeal 2010-006945
Application 10/341,332

2
SUMMARY
Appellants file this appeal under 35 U.S.C. § 134(a) from the
Examiner’s Final Rejection of claims 1-24 and 33-36.1 Specifically, claims
1, 11, 12, 14, 15, 19, and 22-24 stand rejected as unpatentable under 35
U.S.C. § 103(a) as being obvious over the combination of Gustafsson et al.
(US 2001/0044722 A1, November 22, 2001) (“Gustafsson”) and Michael E.
Deisher and Andreas S. Spanias, Speech enhancement using state-based
estimation and sinusoidal modeling, 102(2) J. ACOUST. SCI. AM. 1141 (1997)
(“Deischer”).
Claims 2 and 3 stand rejected as unpatentable under 35 U.S.C. §
103(a) as being obvious over the combination of Gustafsson and Malah et al.
(US 2003/0093279 A1, May 15, 2003) (“Malah”).
Claims 4-6, 16, 34, and 36 stand rejected as unpatentable under 35
U.S.C. § 103(a) as being obvious over the combination of Gustafsson and
Jax et al. (US 2003/0050786 A1, March 13, 2003) (“Jax”).
Claims 7 and 17 stand rejected as unpatentable under 35 U.S.C. §
103(a) as being obvious over the combination of Gustafsson, Asghar et al.
(US 6,418,412 B1, July 9, 2002) (“Asghar”), and Wilson et al. (US
5,323,337, June 21, 1994) (“Wilson”).
Claims 8, 9, 13, and 20-21 stand rejected as unpatentable under 35
U.S.C. § 103(a) as being obvious over the combination of Gustafsson, Jax,
and Gibson et al. (US 6,366,029 B1, January 1, 2003) (“Gibson”).
Claim 18 stands rejected as unpatentable under 35 U.S.C. § 103(a) as
being obvious over the combination of Gustafsson, Asghar, Jax, and Wilson.

1 Claims 25-32 are canceled.
Appeal 2010-006945
Application 10/341,332

3
We have jurisdiction under 35 U.S.C. § 6(b).
We affirm.

NATURE OF THE CLAIMED INVENTION
Appellants’ invention is directed to a method and device for
improving the quality of speech signals transmitted using an audio
bandwidth between 300 Hz and 3.4 kHz. After the received speech signal is
divided into frames, zeros are inserted between samples to double the
sampling frequency. The level of these aliased frequency components is
adjusted using an adaptive algorithm based on the classification of the
speech frame. Sound can be classified into sibilants and non-sibilants, and a
non-sibilant sound can be further classified into a voiced sound and a stop
consonant. The adjustment is based on parameters, such as the number of
zero-crossings and energy distribution, computed from the spectrum of the
up-sampled speech signal between 300 Hz and 3.4 kHz. A new sound with a
bandwidth between 300 Hz and 7.7 kHz is obtained by inverse Fourier
transforming the spectrum of the adjusted, up-sampled sound. Abstract.

GROUPING OF CLAIMS

Because Appellants argue that the Examiner erred for substantially the
same reasons with respect to claims 1-24 and 33-36, we select claim 1 as
representative of this group. App. Br. 37-38. Claim 1 recites:
1. A method comprising:

Appeal 2010-006945
Application 10/341,332

4
classifying speech signals in signal segments of a speech
into a plurality of classes based on at least one signal
characteristic of the speech signals;

converting the speech signals into a plurality of
transformed segments having a speech spectrum in a lower
frequency range and a speech spectrum in a higher frequency
range in a frequency domain, the speech spectrum in the higher
frequency range having a spectral image of the speech spectrum
of the lower frequency range;

modifying the speech spectrum in the high frequency
range in the frequency domain based on the classes for
providing modified transformed segments in said high
frequency range; and

converting the modified transformed segments into
speech data in the time domain.

App. Br. 40.

ISSUES AND ANALYSES
We address each of Appellants’ arguments seriatim, as presented in
Appellants’ Brief.

Issue 1
Appellants argue that the Examiner erred in finding that Gustafsson
teaches or suggests the limitation of claim 1 reciting “classifying speech
signals in signal segments of a speech into a plurality of classes based on at
least one signal characteristic of the speech signals.” App. Br. 21-22. We
therefore address the issue of whether the Examiner so erred.

Appeal 2010-006945
Application 10/341,332

5
Analysis
Appellants admit that Gustafsson teaches or suggests how a speech
signal is sampled into frames and what voiced sound and unvoiced sound
are, but argue that Gustafsson neither teaches nor suggests classifying
speech signals in signal segments of a speech into a plurality of classes
based on at least one signal characteristic of the speech signals. App. Br. 22
(citing Gustafsson, ¶¶ [0062]; [0036]). According to Appellants, Gustafsson
teaches or suggests that whether the speech signal is voiced or unvoiced is
determined by a pitch decision unit 430 based on the error signal 424. App.
Br. 22. Appellants contend that Gustafsson teaches that the pitch decision
unit 430 determines the pitch based on a distance between transients in the
error signal and that, therefore, Gustafsson fails to disclose classifying
speech signals in signal segments of a speech into a plurality of classes
based on at least one signal characteristic of the speech signals. Id. 22-23.
The Examiner responds that Gustafsson teaches or suggests that
speech-sound may be classified into two main categories (i.e., classes):
voiced sound and unvoiced sound (i.e., a signal characteristic). Ans. 5
(citing Gustafsson, ¶ [0062]). The Examiner finds that Gustafsson teaches
or suggests that the processes ensuing between blocks 420-460 of Fig. 4 (see
also Gustafsson, Fig. 6, 620-650; Fig. 8, 830-870) are based upon
Gustafsson’s disclosed “block [frame] of [speech] samples” (i.e., “signal
segments of a speech”). Ans. 20-21 (citing Gustafsson, ¶ [0062]). The
Examiner finds that this teaching includes the process for classifying and
determining voiced/unvoiced sound in blocks 430,630 or 910, which is the
inherent and necessary feature in Gustafsson’s invention because his speech
model-based processes requires frame-by-frame (i.e., signal segment by
Appeal 2010-006945
Application 10/341,332

6
signal segment) operations. Ans. 21. Otherwise, finds the Examiner, the
related parameters (e.g., formant, pitch, etc.) cannot be obtained at all. Id.
The Examiner concludes that a person of ordinary skill in the
contemporaneous art would have readily recognized this inherent feature.
Id.
The Examiner additionally finds that Gustafsson’s teachings with
respect to a vocoder (implying frame-by-frame processing) (citing
Gustafsson, ¶¶ [0053]; [0058]), “segments … represent[ing] an unvoiced
sound” (citing Gustafsson, ¶ [0033]), “segmentation module 820” (citing
Gustafsson, Fig. 8; ¶ [0054]), and “short time stationary” for “the number of
blocks” (citing Gustafsson, ¶¶ [0064]-[0065]), also read on the disputed
limitation. Ans. 21.
We agree with the Examiner. Gustafsson teaches that “the parametric
spectral analysis unit 420 outputs parameters, (i.e., values associated with
the particular model employed therein) descriptive of the received voice
signal” and that the:
[p]itch decision module 430 also determines whether the speech
content of the received signal represents a voiced sound or an
unvoiced sound, and generates a signal indicative thereof. The
decision made by the pitch decision unit 430 regarding the
characteristic of the received signal as being a voiced sound or
an unvoiced sound may be a binary decision or a soft decision
indicating a relative probability of a voiced signal or an un-
voiced signal.2

2 With respect to voiced and unvoiced sounds, Gustafsson teaches that:

Voiced sounds are produced when quasi-periodic bursts of air
are released by the glottis, which is the opening between the
vocal cords. These bursts of air excite the vocal tract, creating a

Appeal 2010-006945
Application 10/341,332

7

The pitch information and a signal indicative of whether
the received signal is a voiced sound or an unvoiced sound are
output from the pitch decision unit 430 to a residual extender
and copy unit 440.

Gustafsson, [0046]-[0048] (emphasis added); see Ans. 21. We agree with
the Examiner that “voiced sound or unvoiced sound” correspond to claim 1’s
“at least one signal characteristic of the speech signals.” Ans. 20-21. We
therefore conclude that the Examiner did not err in finding that Gustafsson
teaches or suggests the limitation of claim 1 reciting “classifying speech
signals in signal segments of a speech into a plurality of classes based on at
least one signal characteristic of the speech signals.”

Issue 2
Appellants argue that the Examiner erred in finding that the
combination of Gustafsson and Deisher teaches or suggests the limitation of
claim 1 reciting “converting the speech signals into a plurality of
transformed segments having a speech spectrum in a lower frequency range
and a speech spectrum in a higher frequency range in a frequency domain,
the speech spectrum in the higher frequency range having a spectral image

voiced sound (i.e., a short “a” (ä) in “car”). By contrast,
unvoiced sounds are created when a steady flow of air is forced
through a constraint in the vocal tract. This constraint is often
near the mouth, causing the air to become turbulent and
generating a noise-like sound (i.e., as “sh” in “she”).

Gustafsson, ¶ [0036].

Appeal 2010-006945
Application 10/341,332

8
of the speech spectrum of the lower frequency range.” App. Br. 23. We
therefore address whether the Examiner so erred.

Analysis
Appellants admit that Gustafsson discloses that the input signal is
upsampled by a factor of 2, and the upsampled signal is analyzed by a
parametric analysis module to determine the formant structure of the
received speech signal either by an autoregressive (“AR”) model in z-
domain or a by sinusoidal mode as described in Deisher. App. Br. 25.
Appellants contend that the results of this analysis are the parameters and an
error signal. Id. Nevertheless, Appellants contend that these teachings and
suggestions of the prior art references do not disclose the disputed limitation
of claim 1. App. Br. 29-32.
With respect to the AR model in z-domain taught or suggested by
Gustafsson, Appellants argue that the Examiner erroneously found that the
AR model in z-domain (transform) corresponds to the equivalent frequency
domain, because the Fourier transform can be viewed as the z-transform of
the sequence evaluated on the unit circle, owing to the two transforms
inherently having the relationship X(z)|z=e jω =X(ω). App. Br. 29.
Appellants admit that there is a relationship between the Fourier
transform and z-transform, but argue that the essential difference between
Gustafsson and the present invention is that Gustafsson applies the z-
transform only to the envelope of the spectrum, and so the result is not a
“speech spectrum.” Id. Appellants assert that the Fourier transformation
provides the frequency domain representation of the entire signal being
transformed, whereas the AR model taught by Gustafsson, when applied to
Appeal 2010-006945
Application 10/341,332

9
the same signal, yields two separate items: the residual signal and the
envelope. Id. Of the two items, only the parameters can be represented in
the z-domain. Id. (citing Gustafsson, Eq. 2). The residual signal is a time-
domain entity which is used to provide information on the peaks and/or peak
locations in the upsampled input signal in the narrowband. App Br. 29.
Appellants therefore contend that an AR-model in z-domain cannot be used
to convert an upsampled input speech signal into transformed segments in
the frequency domain having a spectrum in the lower frequency range and a
spectrum in the higher frequency range. Id.
With respect to the sinusoidal model taught by Deisher and explicitly
incorporated by reference in Gustafsson,3 Appellants argue that a sinusoidal
model is one-half of a Fourier transform, because it lacks the cosine function
for the same harmonic frequency. App. Br. 30. Appellants contend that
since a sinusoidal model is only one half of a Fourier transform; it cannot be
used to convert an upsampled input speech signal into transformed segments
in the frequency domain having a spectrum in the lower frequency range and
a spectrum in a higher frequency range. Id.
In summary, Appellants argue that the AR-model in z-domain is
different from a fast Fourier transform (“FFT”) in that an FFT provides the
frequency domain representation of the entire signal being transformed,
whereas an autoregressive (AR) model applied to the same signal yields two
separate items: the residual signal and the envelope, of which the only the
parameters can be represented in the z-domain. App. Br. 34. Furthermore,
because the sinusoidal mode lacks the corresponding cosine terms, the

3 See Gustafsson, ¶ [0046].
Appeal 2010-006945
Application 10/341,332

10
sinusoidal model does not convert the upsampled speech signal into
transformed segments in the frequency domain as an FFT does. Id.
Consequently, Appellants argue, combination of the two models still cannot
be used to convert an upsampled speech signal into transformed segments in
the frequency domain. App. Br. 34-35.
Finally, Appellants argue that, even if the upsampled speech signal is
converted into transformed segments in the frequency domain, this output of
the conversion would completely change the principles of operation taught
or suggested by Gustafsson. App. Br. 35. According to Appellants, if the
output of Gustafsson’s parametric spectral analysis unit 420 is transformed
segments in the frequency domain, most, if not all, of the remaining part of
the speech synthesis systems as disclosed in Gustafsson will be inoperable
because (a) the pitch decision unit 430, the residual extender/copy unit 440,
and the residual modifier can only be used to process a time-domain entity
such as an error signal 424; (b) the synthesis filter 450 and the artificial
formant determining unit 470 relies on the parameters which are values
associated with a certain AR or sinusoidal model; and (c) the new formant
determining unit 850 relies on the parameters as calculated by the linear
prediction module 840. Id. Appellants therefore assert that Gustafsson does
not teach or suggest converting the input signal into transformed segments in
the frequency domain having a spectrum in a lower frequency range and a
spectrum in the higher frequency range. App. Br. 36.
The Examiner responds that Gustafsson teaches the spectral analysis
module using the alternative sinusoidal model, which inherently processes
the signal in DFT (discrete Fourier transform) (or FFT) domain (i.e.,
frequency domain). The Examiner finds that an inherent feature of the
Appeal 2010-006945
Application 10/341,332

11
sinusoidal model is that “the short speech segments” are
analyzed/synthesized (including enhanced or modified) with DFT/FFT (i.e.,
transformed segments) in a frequency-domain. Ans. 21 (citing Deisher, §§ I,
III; Fig. 1). The Examiner finds, contra Appellants’ argument, that Equation
(1) of Deisher, which describes the ith sample of speech within a 20 to 30
ms block, is not just half of the Fourier transform, because the existence of
the variable phases implies presence of sines, which would explicitly appear
by using trig identity for the cosine of the sum of two angles, which reads on
the disputed limitation. Ans. 22 (citing Deisher, § 1).
The Examiner further finds that, in another embodiment, Gustafsson
teaches a spectral analysis module using the AR model, wherein the speech
model parameters includes envelope spectrum parameters (e.g., ak
representing the formant) processed in z-domain and excitation spectrum
parameters (such as pitch, peak/harmonic numbers or locations) processed
via Fourier transform (i.e. frequency domain). Ans. 22. The Examiner finds
that, since Gustafsson’s speech model is processed based on frame-by-frame
sections, (i.e., segments), the processed frames/blocks combining the speech
model signal and residual/error signal in or after blocks 420, 620, or 830
(Figs. 4, 6, or 8) read on the claimed “transformed segments.” The
Examiner finds that, since the determined parameters, such as envelope
spectrum parameters and excitation spectrum parameters, represent each of
the processed frames of speech signal based on the speech model, the
combination of them as a whole corresponds to the claimed transformed
segments (i.e., transformed to speech model based parametric segments).
Ans. 22.
Appeal 2010-006945
Application 10/341,332

12
The Examiner also finds that the claimed “speech spectrum” can be
reasonably interpreted as the entire speech spectrum, amplitude or power
spectrum, phase spectrum, envelope spectrum, excitation spectrum, or
possible combinations thereof, because the specification and claim do not
clearly or specifically define the term “speech spectrum.” Ans. 22-23.
Moreover, finds the Examiner, since Fourier transform (X(ω)) and z-
transform (X(z)) related by ( X(z)|z=e jω = X(ω)), the z-domain processes
taught or suggested by Gustafsson are reasonably interpreted as being
equivalently in frequency domain and one of ordinary skill in the art would
have readily recognized that the processes in the two transforms are
interchangeable and would produce a predictable result. Id.
We are not persuaded by Appellants’ arguments. As an initial matter,
we observe that the language of claim 1 does not explicitly require the use of
a fast Fourier transform to produce transformed segments in the frequency
domain, as Appellants appear to imply. App. Br. 34. Although the use of an
FFT is described in Appellants’ Specification (see, e.g., Spec., Fig. 1),
limitations of an embodiment in the Specification may not be read into the
claims. See Phillips v. AWH Corp., 415 F.3d 1303, 1323 (Fed. Cir. 2005).
We agree with the Examiner that Gustafsson teaches or suggests
“converting the speech signals into a plurality of transformed segments
having a speech spectrum in a lower frequency range and a speech spectrum
in a higher frequency range in a frequency domain, the speech spectrum in
the higher frequency range having a spectral image of the speech spectrum
of the lower frequency range.” Specifically, we agree with the Examiner
that the AR model and the sinusoidal model taught or suggested by
Gustafsson (the latter being incorporated by reference from Deisher)
Appeal 2010-006945
Application 10/341,332

13
correspond to the “transformed segments having a speech spectrum … in a
frequency domain.” Ans. 21. Deisher teaches or suggests that the sinusoidal
model represents speech over short time intervals by a finite set of sinusoids
that capture most of the signal energy; the model’s amplitudes Ak and ϕk are
typically measured from a high resolution DFT which are in the frequency
domain. Deisher, § I; see Ans. 22.
Moreover, although we agree with Appellants that the AR model
provides two results, an envelope of the spectrum in the frequency domain
and a residual, we find that this result corresponds to the limitation reciting
“transformed segments having a speech spectrum … in a frequency
domain,” because although an FFT is taught by Appellants’ Specification,
the language of the claims requires only that the segment be “transformed
…having a speech spectrum … in the frequency domain.” Gustafsson
teaches that the AR model meets this criterion. See Gustafsson, Figs. 13-15;
Ans. 7.
Nor are we persuaded by Appellants’ argument that when the
upsampled speech signal is converted into transformed segments in the
frequency domain, the output of the conversion would completely change
the principles of operation taught or suggested by Gustafsson. App. Br. 35.
“The test for obviousness is not whether the features of a secondary
reference may be bodily incorporated into the structure of the primary
reference.” In re Keller, 642 F.2d 413, 425 (CCPA 1981). Rather, the
obviousness inquiry requires a finding that the combination of known
elements was obvious to a person with ordinary skill in the art. See In re
Chevalier, No. 2012-1254, 2013 WL 57873, at *3 (Fed. Cir. Jan. 7, 2013);
see also In re Sneed, 710 F.2d 1544, 1550 (Fed. Cir. 1983) (“[I]t is not
Appeal 2010-006945
Application 10/341,332

14
necessary that the inventions of the references be physically combinable to
render obvious the invention under review”); and In re Nievelt, 482 F.2d 965
(CCPA 1973) (“Combining the teachings of references does not involve an
ability to combine their specific structures”). We conclude that these
elements of the disputed limitation would, through the teachings and
suggestions of Gustafsson and Deisher, have been known to a person of
ordinary skill in the contemporary art.
We consequently conclude that the Examiner did not err in finding
that the combined prior art references teach or suggest the limitation of
claim 1 reciting “converting the speech signals into a plurality of
transformed segments having a speech spectrum in a lower frequency range
and a speech spectrum in a higher frequency range in a frequency domain,
the speech spectrum in the higher frequency range having a spectral image
of the speech spectrum of the lower frequency range.”

Issue 3
Appellants argue that the Examiner erred in finding that the prior art
references teach or suggest the limitation of claim 1 reciting “modifying the
speech spectrum in the high frequency range in the frequency domain.”
App. Br. 36. We therefore address the issue of whether the Examiner so
erred.

Analysis
Appellants argue that Gustafsson teaches or suggests how a synthetic
formant can be added to the synthesized speech signal in the higher
frequency range. App. Br. 31. However, argue Appellants, when an AR
Appeal 2010-006945
Application 10/341,332

15
model in z-domain is used, the error signal and the parameters are conveyed
to an inverse filter for speech synthesis (synthesis filter 450), the synthesized
speech will have the same formants if the inverse filter uses the same filter
coefficients. Id. Appellants argue that the inverse or synthesis filter can be
an AR model in z-domain or a sinusoidal model taught or suggested by
Gustafsson. Id. Appellants contend that parameters and error signal are not
equivalent to transformed segments in the frequency domain having a
speech spectrum in the lower frequency range and a speech spectrum in the
higher frequency range, the inverse or synthesis filter does not modify the
transformed segments in the frequency domain. Id.
The Examiner responds that Gustafsson teaches or suggests that, when
employing the sinusoidal model the short speech segments after DFT/FFT
(becoming transformed segments) are analyzed and/or synthesized,
including enhanced (i.e., becoming modified transformed segments), such as
“selection of harmonic components” and “harmonic reproduction” in
“frequency-domain.” Ans. 24 (citing Deisher, Abstract; §§ I, III; see also
Fig. 1(“enhancement” processes between blocks “FFT” and “inversed
FFT”)). The Examiner also finds that the parameters taught or suggested by
Gustafsson include modified/added envelope spectrum parameters (e.g.,
formant in Fig. 8 and ¶¶ [0078]-[0084]) and/or excitation spectrum
parameters (e.g., peaks or harmonics for residual or error signal in Figs. 5; 7;
9; ¶¶ [0050]-[0051]) based on voiced/unvoiced information (Figs. 4; 6; 9)
Ans. 24-25. The Examiner finds that these teaching and suggestions of
Gustafsson reasonably read on the claim language of “modifying the speech
spectrum.” Ans. 25.
Appeal 2010-006945
Application 10/341,332

16
Furthermore, finds the Examiner, in the embodiment of Gustafsson
teaching use of the AR model in z-domain, the modification of the envelope
spectrum (e.g., formant in Fig. 8; ¶¶ [0078]-[0084]) and/or
excitation spectrum (e.g., peaks or harmonics for residual/error signal in
Figs. 5, 7, and 9 and ¶¶ [0050]-[0051]) is based on voiced/unvoiced
information (Figs. 4, 6, 9) for synthesizing wideband speech. Ans. 25 (citing
Gustafsson, ¶¶ [0046]-[0048]). The Examiner finds that one of ordinary
skill in the contemporaneous art would have recognized that the AR process
in z-domain (transform) would be interpreted as being equivalently in the
frequency domain in a broader sense under the inherent relationship, as
related supra. Id. The Examiner therefore finds that the parameters taught
or suggested by Gustafsson, including modified/added envelope spectrum
parameters (such as formant) and excitation spectrum parameters (such as
peaks or harmonics for residual or error signal) represent an extended speech
frame, and can be reasonably interpreted as reading on modified transformed
segments. Ans. 25.
We are not persuaded by Appellants’ arguments. We have related
supra why we agree with the Examiner’s findings that Gustafsson teaches or
suggests transformed segments in the frequency domain. Moreover, we find
that Gustafsson teaches supplementing the envelope of the speech power
spectrum (transfer function) with synthetic formants and/or high-pass
filtered noise. Gustafsson, Figs. 13-15; ¶¶ [0084]-[0086]; see Ans. 7.
Further, we agree with the Examiner that Gustafsson teaches or suggests that
“modifier 930 creates harmonic peaks in the upper frequency band (in
frequency domain) by copying parts of the lower band residual signal to the
higher band” and “create the peaks corresponding to the high order
Appeal 2010-006945
Application 10/341,332

17
harmonics in the upper frequency band.” Ans. 7 (citing Gustafsson,
¶¶ [0059]-[0061], Figs. 9, 13). We find that these teachings and suggestions
of the prior art citations correspond to the language of claim 1 reciting
“modifying the speech spectrum in the high frequency range in the
frequency domain” and we therefore conclude that the Examiner did not err
in concluding that that limitation would be obvious over the prior art
references a person of ordinary skill in the contemporaneous art.

DECISION

The Examiner’s rejection of claims 1-24 and 33-36 as unpatentable
under 35 U.S.C. 103(a) is affirmed.

TIME PERIOD FOR RESPONSE
No time period for taking any subsequent action in connection with
this appeal may be extended under 37 C.F.R. § 1.136(a) (1)(iv) .

AFFIRMED

msc