Electrical audio signal processing systems and devices – Including frequency control – Having automatic equalizer circuit
Reexamination Certificate
1999-04-27
2004-02-24
Lee, Ping (Department: 2644)
Electrical audio signal processing systems and devices
Including frequency control
Having automatic equalizer circuit
C381S094300
Reexamination Certificate
active
06697492
ABSTRACT:
FIELD OF THE INVENTION
The present invention pertains to a method for reproducing true source sound from acoustic speakers using digital signal processing (DSP) technology.
BACKGROUND OF THE INVENTION
With the unprecedented increase in the use of multimedia information and the introduction of high-definition digital TV broadcasting, high-quality acoustic sound has become an essential element for producing a more realistic feel over what was produced in the past.
Acoustic waves generated by conventional acoustic speakers contain significant distortion because of the physical limitations on the mechanical structure of the speaker. Regardless of careful processing in-studio, or recording high-quality digital data using media such as digital audio tapes (DAT) or compact disks (CD), actual acoustic waves generated by conventional acoustic speakers are normally very different from the original source sounds and from high-precision reproductions of raw source sounds.
FIG. 1
a
shows a typical conventional acoustic system, and
FIG. 1
b
shows typical frequency characteristics. In these FIGS., source sounds recorded on sound source
100
, such as a compact disk, are converted to acoustic output by acoustic speaker
104
via amplifier
101
, control amplifier
102
, and power amplifier
103
. Distortion in the reproduced audio is produced in the various stages, that is, the amplifiers
101
,
102
, and
103
, and acoustic speaker
104
.
FIGS. 2
a
and
b
represent the difference between original source sound and acoustic output measured using a reference microphone.
FIG. 3
shows an acoustic reproduction system defined as a signal processing box provided with input/output characteristics H(&ohgr;). This acoustic reproduction system can be viewed as unknown box
110
that has known (measurable) input/output characteristics (distortion).
So this creates the problem of whether an additional signal processing box or filter that has inverse characteristics H
−1
(&ohgr;) can be placed somewhere in this system, and if the original high-quality source sound can be restored in the output from acoustic speakers.
A filter is used as the measure for solving the present problem.
FIG. 4
shows an acoustic speaker system that uses an inverse filter for high-quality acoustic reproduction. As is clear from the FIG., filter
115
with inverse characteristics H
−1
(&ohgr;) is inserted between power amplifier
103
and acoustic speaker
104
.
In the past, some research was carried out to achieve an ideal system that has flat amplification and linear phase characteristics. In theory, there are two approaches to this. In short, there is a time domain algorithm and a frequency domain algorithm. All approaches up to now have used only time domain algorithms. With this approach, the inverse filter coefficient is found with the method shown in
FIG. 5
, and filter coefficient w
IK
is updated adaptively so that the difference between original input audio signal
120
and output
121
from an acoustic is minimized. Filter coefficient calculation circuit
122
in
FIG. 5
illustrates calculation of filter coefficient w
IK
with least squares method (LMS) algorithm
123
and the coefficient found here is given to filtering processing circuit
124
.
The Kalman filtering theory is well-known. With this, updating of filter coefficient w
IK
is orthogonal to filtering error e
k
. This error e
k
is the difference between output delayed by delay circuit
125
and output filtered by filtering processing circuit
126
of filter coefficient calculation circuit
122
. Delay z
−&Dgr;
is required to compensate for filtering delay. Expressed mathematically, filter coefficient updating is given by the equation below.
w
(
k
+
1
)
=
w
(
k
)
+
α
x
(
k
)
T
x
(
k
)
e
(
k
)
x
(
k
)
Here, filter coefficient w
IK
and input signal x
K
are both given in the form of specific vectors.
The filter coefficient can be found by using either actual audio signals or reference white noise as input. Use of the latter has been shown to give a more accurate filter coefficient. However, in a dynamic environment, acoustic speaker characteristics must be compensated for adaptively, and actual signals must be used as input and the filter coefficient must be calculated adaptively.
A conventional approach of this type that obtains filter coefficients in time domain is simple, so it is often applied to actual practice. However, it has the important problems described below.
FIG. 6
illustrates typical actual acoustic speaker characteristics. It shows equalizable frequency range and unequalizable frequency range.
Due to physical limitations of acoustic speakers, a general characteristic of acoustic speakers, as shown in
FIG. 6
, is that they undergo significant attenuation in very low frequency bands and in very high frequency bands. This phenomenon is caused by physical limitations in speaker structure, and since it is generally a nonlinear property, there is the concern that trying to restore these characteristics will lead to acute deterioration in all speaker characteristics.
A conventional approach to recover acoustic signals in time domain tries to extract filter coefficients independently from this phenomenon. For this reason, rather than being satisfactory, it has led to gradual unexpected deterioration in acoustic speaker characteristics.
SUMMARY OF THE INVENTION
The present invention, in order to solve the aforementioned problems in the prior art, uses an audio signal processing method where the entire frequency band of input audio signals is divided into multiple sub-bands, equalizable sub-bands are recognized from each sub-band and at the same time a filter coefficient is found by comparing the equalizable sub-bands and corresponding sub-bands from output audio signals, and frequency convolution calculates a filter coefficient for the calculated sub-band, and input audio signals are processed based on this convolution.
Also, the audio signal processing system associated with the present invention has a sub-band analysis filter bank that divides the entire frequency band of input audio signals into multiple sub-bands, a filter coefficient calculating circuit that identifies equalizable sub-bands from each sub-band and at the same time calculates filter coefficients by comparing equalizable sub-bands and corresponding sub-bands from the output audio signals, and a processing circuit that performs frequency convolution on calculated filter coefficients for equalizable sub-bands and processes input audio signals based on this convolution.
Also, the digital signal processing device associated with the present invention is equipped with a data memory for storing data, a program memory for storing command programs, a multiplier, and a control unit. The aforementioned control unit enables control of writing of data to the aforementioned data memory and control of the aforementioned multiplier in response to command programs stored in the aforementioned program memory. The aforementioned data memory stores data for multiple sub-bands obtained by dividing input audio signals into multiple frequency bands and data for multiple sub-bands obtained by dividing reference audio signals into multiple frequency bands. The aforementioned program memory stores command programs for multiplying the ratio of the data for each sub-band of the aforementioned audio signals and the data for each sub-band of the aforementioned reference audio signals by the aforementioned filter coefficient in order to correct filter coefficients found from each sub-band for the aforementioned input audio signals and from each sub-band for the aforementioned reference audio signals, and for storing the results of said multiplication. The aforementioned control unit enables multiplication of the aforementioned ratio and the aforementioned filter coefficient by the aforementioned multiplier and stores the corrected filter coefficient in the aforementioned data memory.
Here “equalize” means to restore distorted signal
Higa Yoshito
Yamaguchi Hirohisa
Brady III W. James
Lee Ping
Marshall, Jr. Robert D.
Telecky , Jr. Frederick J.
Texas Instruments Incorporated
LandOfFree
Digital signal processing acoustic speaker 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 Digital signal processing acoustic speaker system, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Digital signal processing acoustic speaker system will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3352723