Coded data generation or conversion – Analog to or from digital conversion – Analog to digital conversion
Reexamination Certificate
2001-11-28
2002-07-02
Young, Brian (Department: 2819)
Coded data generation or conversion
Analog to or from digital conversion
Analog to digital conversion
Reexamination Certificate
active
06414620
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a signal processing system, and in particular, to a signal processing system having a voice processing function such as, for example, a voice recognition function or a voice synthesis function, which is required to be performed in real-time in mobile electronic devices including cellular phones.
2. Description of the Related Art
FIG. 6
is a block diagram illustrating a conventional voice processing system
600
.
The voice processing system
600
includes a CODEC
601
, a DSP (digital signal processor)
602
, non-volatile MROMs (mask ROMs)
603
and
606
, a synchronous DRAM
604
, and a microprocessor (CPU)
605
. The DSP
602
is connected to the non-volatile MROM
606
, the synchronous DRAM
604
, and the microprocessor
605
via a bus
607
.
The CODEC
601
includes an A/D converter for converting an analog voice signal input from an external device into a digital signal, and a D/A converter for converting a digital signal obtained by processing performed in the voice processing system
600
into an analog signal. The DSP
602
processes the digital signal obtained by the A/D converter in the CODEC
601
. The non-volatile MROM
603
stores an acoustic model used for extracting a feature amount of the input analog voice signal. The non-volatile MROM
606
stores, for example, a voice recognition program, a voice synthesis program, and dictionary data, which are required for processing performed by the DSP
602
. The synchronous DRAM
604
is a volatile memory for temporarily storing data processed by the DSP
602
. The microprocessor
605
transfers the programs stored in the non-volatile MROM
606
to the DSP
602
and controls the DSP
602
to execute the programs.
The voice processing system
600
shown in
FIG. 6
performs voice recognition as described below in steps S
01
through S
04
.
Step S
01
: An analog voice signal input from an external device is converted into a digital signal by the A/D converter in the CODEC
601
.
Step S
02
: The microprocessor
605
transfers the voice recognition program stored in the non-volatile MROM
606
to a program memory in the DSP
602
. Alternatively, the microprocessor
605
transfers the voice recognition program, the dictionary data and the like stored in the non-volatile MROM
606
to the synchronous DRAM
604
and controls the DSP
602
to directly access the synchronous DRAM
604
and to read the voice recognition program, the dictionary data and the like transferred.
Step S
03
: The DSP
602
extracts a feature amount of the digital signal obtained in step S
01
in accordance with the voice recognition program. For extracting the feature amount, the DSP
602
is controlled by the microprocessor
605
to read the acoustic model from the non-volatile MROM
603
while comparing the digital signal with the acoustic model stored in the non-volatile MROM
603
. The obtained feature amount data is temporarily stored in the synchronous DRAM
604
via the bus
607
.
Step S
04
: The DSP
602
identifies input voice from the feature amount data obtained in step S
03
. The DSP
602
reads the dictionary data stored in the non-volatile MROM
606
(or the dictionary data transferred to the synchronous DRAM
604
) via the bus
607
. Then, the DSP
602
reads the feature amount data temporarily stored in the synchronous DRAM
604
via the bus
607
. The DSP
602
compares the dictionary data with the feature amount data read from the synchronous DRAM
604
. The DSP
602
then temporarily stores intermediate data obtained by the comparison in the synchronous DRAM
604
via the bus
607
. Since the bus
607
is occupied for storing the intermediate data in the synchronous DRAM
604
, the comparison between the dictionary data and the subsequent portion of the feature amount data is interrupted while the intermediate data is being stored in the synchronous DRAM
604
. In this manner, the entire feature amount data and the dictionary data are compared, so as to identify the input voice.
The voice processing system
600
performs voice synthesis in a manner similar to voice recognition as described below.
The voice synthesis program stored in the non-volatile MROM
606
is transferred to the program memory in the DSP
602
by an instruction from the microprocessor
605
. In accordance with the voice synthesis program, the DSP
602
synthesizes voice from text data to be synthesized, and the D/A converter in the CODEC
601
converts the obtained digital voice signal into an analog signal. The analog signal is output through an external speaker.
The conventional voice processing system
600
shown in
FIG. 6
typically uses the non-volatile MROMs
603
and
606
as non-volatile memories. An MROM already has information written therein when being produced, and the information cannot be rewritten by the user. An MROM is used for the reasons that (1) a voice recognition program, a voice synthesis program, dictionary data, an acoustic model and the like which are required for voice recognition or voice synthesis need not be rewritten, and (2) an MROM is inexpensive and cost-effective.
However, in order to improve the voice recognition accuracy, the acoustic model needs to be optimized by rewriting the data. The following two types of voice recognition systems, for example, use a rewritable non-volatile memory.
FIG. 7
is a block diagram illustrating a conventional voice recognition system
700
used for car navigation, which is disclosed in Japanese Laid-Open Publication No. 10-282987.
The voice recognition system
700
includes a microphone
701
for taking in voice, a dictionary switching section
703
for selecting a dictionary or switching one dictionary to another in accordance with dictionary switching information
702
, a non-volatile ROM
704
storing a plurality of dictionary data units, a volatile RAM
705
for storing a dictionary data unit transferred from the non-volatile ROM
704
, a voice analysis section
706
for performing pre-processing such as, for example, noise processing or voice analysis, a voice recognition section
707
for performing voice recognition, and an acoustic model section
708
storing an acoustic model to be read when the voice recognition section
707
performs voice recognition.
The voice recognition section
707
outputs a voice recognition result by a signal
709
, and also feeds the voice recognition result back to the dictionary switching section
703
by a signal
710
.
The non-volatile ROM
704
is a rewritable memory such as, for example, a flash memory. The plurality of dictionary data units stored in the non-volatile ROM
704
are transferred to the volatile RAM
705
when necessary. The voice recognition section
707
accesses the volatile RAM
705
storing the dictionary data unit transferred from the non-volatile ROM
704
.
The voice processing system
700
. stores the dictionary data units in the rewritable non-volatile ROM
704
, and therefore improves the voice recognition accuracy in a car navigation system which needs to have a large vocabulary. Since the voice recognition section
707
accesses the volatile RAM
705
storing the dictionary data unit transferred thereto, high-speed data read is realized due to the characteristics of the RAM (random access memory) and thus the response speed of the voice processing system
700
is increased.
FIG. 8
is a block diagram illustrating a conventional voice processing system
800
used in a cellular phone, which is disclosed in Japanese Laid-Open Publication No. 11-345194.
The voice processing system
800
includes a CPU
801
for controlling elements of the voice processing system
800
, a DSP
802
including a volatile RAM
803
, and a non-volatile ROM
804
storing a voice CODEC, a program and the like. These elements are connected with each other as shown in
FIG. 8
via a bus
805
.
The non-volatile ROM
804
is a rewritable memory such as, for example, a flash memory. The program stored in the non-volatile ROM
804
is transferred to the volatile RAM
803
i
Kawai Tomoyuki
Maeda Kengo
Nguyen John
Sharp Kabushiki Kaisha
Young Brian
LandOfFree
Sound processing system does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Sound processing system, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Sound processing system will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2819264