Ex Parte Kiuchi et al

Board of Patent Appeals and Interferences

Mar 29, 2012

10730767 (B.P.A.I. Mar. 29, 2012)

UNITED STATES PATENT AND TRADEMARK OFFICE
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.
10/730,767 12/08/2003 Shingo Kiuchi 9333-361 3437
74989 7590 03/30/2012
ALPINE/BHGL
P.O. Box 10395
Chicago, IL 60610
EXAMINER
WOZNIAK, JAMES S
ART UNIT PAPER NUMBER
2626
MAIL DATE DELIVERY MODE
03/30/2012 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 BOARD OF PATENT APPEALS
AND INTERFERENCES
____________

Ex parte SHINGO KIUCHI and NOZOMU SAITO
____________

Appeal 2009-013722
Application 10/730,767
Technology Center 2600
____________

Before ELENI MANTIS MERCADER, CARL W. WHITEHEAD, JR. and
BRADLEY W. BAUMEISTER, Administrative Patent Judges.

WHITEHEAD, JR., Administrative Patent Judge.

DECISION ON APPEAL
Appeal 2009-013722
Application 10/730,767

2
STATEMENT OF THE CASE
Appellants appeal under 35 U.S.C. § 134 (a) from a final rejection of
claims 1, 3-8, 12-15, 17-18, and 20. Appeal Brief 2 and 6. We have
jurisdiction under 35 U.S.C. § 6(b) (2002).
We reverse.

Introduction
The invention relates to “a device and method for improving speech
recognition performance in a noisy environment.” Appeal Brief 3.

Exemplary Claim
Exemplary independent claim 1 under appeal reads as follows:
1. A method for use with a speech recognition device
for improving speech recognition performance, said method
comprising:
identifying a start position of a speech region of speech
data for which speech recognition is to be performed;
generating, from said speech data for which speech
recognition is to be performed, a plurality of pieces of speech
data including said speech region and a varying period of a
preceding non-speech region, where start positions of non-
speech regions differ for the plurality of pieces of speech data;
performing speech recognition using each of said pieces
of speech data to obtain a plurality of recognized results; and
identifying a most numerous recognized result from
among the plurality of obtained recognized results;
wherein, by sequentially shifting the start position of said
non-speech region from the start position of the speech region
back to a position preceding by a predetermined time, a
plurality of pieces of speech data whose start positions of non-
Appeal 2009-013722
Application 10/730,767

3
speech regions differ are generated from said speech data for
which speech recognition is to be performed.

Rejection on Appeal
Claims 1, 3-8, 12-15, 17-18, and 20 stand rejected under 35 U.S.C.
§ 103 (a) as being unpatentable over Fujii et al (U.S. Patent Number
4,885,791(filed October 20, 1986) (issued December 5, 1989)), Keiller (U.S.
Patent Number 6,975,993 B1(filed May 2, 2000) (issued December 13,
2005)), and Bi (U.S. Patent Number 6,324,509 B1(filed February 8,
1999)(issued November 27, 2001)). Answer 3-6.

ANALYSIS
The Examiner contends Appellants’ invention is not concerned with
determining an actual starting point of speech because the instant invention
involves “determining a plurality of starting points in a supposed non-speech
region in order to cope with a high noise level in a recognition
environment.” Answer 7 (citing Specification 3, paragraph 0010]).
Examiner surmises that both the instant invention and Fujii are both similar
in that regard. Id. Examiner concludes the Appellants’ claims are silent on
determining an actual starting point of speech because claim 1 merely sets
forth identifying a start position, thus only requiring a general start position.
Id. at 7-8.
Appellants argue the following:
Characterizations aside, the fact is that Appellants’ claims recite
“identifying a start position of a speech region of speech data for
which speech recognition is to be performed,” period. Nothing could
Appeal 2009-013722
Application 10/730,767

4
be plainer or clearer: a start position of the speech region is identified
-- not some vague “general” start position as asserted by the
Examiner. Thus, contrary to the Examiner's assertion, Appellants’
invention does involve determining a starting point of a speech region
of speech data, and the claims expressly recite this.
Reply Brief 2 (underline omitted).
We find Appellants’ arguments persuasive. The claims indicate a
starting position is identified, thus, leaving the interpretation of the claims
unambiguous .
The Examiner finds:
Fujii, Keiller, and Bi are analogous art because they are from a
similar field of endeavor in speech recognition systems. Thus, it
would have been obvious to a person of ordinary skill in the art, at the
time of invention, to modify the teachings of Fujii in view of Keiller
with the concept of backwards searching (shifting) taught by Bi in
order to provide a well-known means of achieving the multiple speech
data periods in Fujii that can be easily implemented in a real-time
processor (Bi, Col. 5, Lines 24-30).
Answer 5.
Further, the Examiner concludes that since the claims do not require
determining an “actual starting point of speech” and “the combination of
Fujii and Bi teaches ‘identifying a starting position,’” The Examiner’s
reasoning is not convincing. Answer 8.
Appellants argue the passage (column 5, lines 13-30) of Bi relied
upon by the Examiner is not applicable because the passage “merely notes
that the signal data are [sic] stored in a buffer because the processor
performs real-time processing but must be able to look back a certain
number of speech frames.” Appeal Brief 9.
Appellants further argue:
Appeal 2009-013722
Application 10/730,767

5
In particular, Bi does not describe sequentially shifting
backwards to obtain a plurality of starting points. Bi uses a first
signal-to-noise ratio (SNR) threshold value to identify a “first starting
point” and a “first ending point” of calculation that are not endpoints
of the speech data but are instead only interim calculation points
(“PRE_START” and “PRE END”). Then, Bi uses a second, smaller
SNR threshold value to determine the “actual” starting and ending
points of the speech data (e.g., Abstract, col. 4, lines 37-57; col. 6,
lines 15-39; col. 7, lines 24-30).
Bi notes that in conventional voice recognition devices, the
endpoint detector relies upon a single SNR threshold to determine the
endpoints of a piece of speech. However, setting the SNR threshold
too low may make the device too sensitive to background noise,
whereas setting the SNR threshold too high may miss part of the
beginning or ending of speech (col. 2, lines 21-37). Bi, in contrast,
“uses multiple, adaptive SNR thresholds to accurately detect the
endpoints of speech in the presence of background noise” (col. 2, lines
42-44). Thus, Bi explicitly states that it only determines one actual
starting point and one actual ending point.
Bi does not “consider multiple starting points” or disclose
“different possible starting points” as the Office Action and the
Advisory Action assert. Further, the beginning point of the Bi’s
process is not the start of the speech period (as in Applicants’
invention), but rather is simply a point where the SNR reaches a first
arbitrary threshold value.
Appeal Brief 9 (underlines omitted).
We find Appellants’ arguments persuasive. As we stated above, the
claims indicate a starting position is identified, thus leaving the
interpretation of the claims unambiguous. The Examiner cannot ignore the
limitation of identifying a start position of a speech region of speech data.
Further, Bi does not address the deficiencies of Fujii as noted by the
Examiner and as argued by the Appellants. Therefore, we do not sustain the
obviousness rejection of claims 1, 3-8, 12-15, 17-18, and 20.

Appeal 2009-013722
Application 10/730,767

6
DECISION
The rejection of claims 1, 3-8, 12-15, 17-18, and 20 under 35 U.S.C.
§ 103 is reversed.

REVERSED

llw