Coded data generation or conversion – Digital code to digital code converters – Unnecessary data suppression
Reexamination Certificate
1999-11-03
2001-07-03
Young, Brian (Department: 2819)
Coded data generation or conversion
Digital code to digital code converters
Unnecessary data suppression
Reexamination Certificate
active
06255968
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to a data compression method and apparatus and, more particularly, to a data compression method and apparatus for compressing and outputting an analog input signal.
Conventionally, when one- or two-dimensional analog data such as audio data, mechanical vibration waveform, and image are to be compressed, the data are temporarily sampled on the time axis, quantized into digital data by an A/D converter, and compressed by digital arithmetic processing.
Digital data (original data) are compressed and reconstructed by a loss-less compression method of completely reconstructing the original data or a lossy compression method by which part of the original data is lost. The present invention approximates sampled data by another waveform, so that the lossy compression method will be explained.
As shown in
FIGS. 12A and 12B
, conventional compression processes are roughly classified into a method using orthogonal transform and a method using function approximation.
According to the method using orthogonal transform, as shown in
FIG. 12A
, an analog input signal is sampled and quantized by a sampling unit
91
and quantization unit
92
, respectively. The original data (digital data) output from the quantization unit
92
is divided into small sections by an orthogonal transform unit
93
A.
Each section is decomposed into frequency space elements by orthogonal transform such as Fourier transform, DCT (Discrete Cosine Transform), or Walsh transform. This decreases the correlation degree between respective sample point on the original data.
The magnitude of each frequency element after orthogonal transform is encoded by a parameter encoder
94
using a code table, thereby obtaining compressed data.
According to the method using function approximation, as shown in
FIG. 12B
, the orthogonal transform unit
93
A is replaced by a function approximation unit
93
B. The feature of each section of original data (digital data) output from the quantization unit
92
is approximated by any function.
The approximated bit data (digital data) is encoded by the parameter encoder
94
using a variable-length coding table of Huffman coding or the like.
As orthogonal transform, there is proposed a method using orthogonal transform such as Slant, Haar, or Wavelet, in addition to the above transform methods. An appropriate orthogonal transform method is used in consideration of various factors such as the property and simple calculation of original data, simple circuit arrangement, and simple high-speed processing.
Further, the following method is also used. Quantized digital data is segmented into small data sections in the original data space (real space), each section is approximated using a simple algebraic curve (e.g., a straight line which approximates each data section), and the algebraic function coefficient is compressed and stored.
To the contrary, methods of reconstructing the original data waveform from compressed data can be roughly classified into a method using orthogonal transform and a method using function approximation.
According to the compression method shown in
FIG. 13A
using orthogonal transform, a parameter decoder
95
calculates orthogonal transform coefficients from compressed data using a code table.
An inverse orthogonal transform unit
96
A performs inverse orthogonal transform using these coefficients to calculate approximate data of original data for each section. A D/A converter
97
converts the data into an analog signal to obtain an analog output signal approximated to the original data through a low-pass filter
98
.
According to the method using function approximation, as shown in
FIG. 13B
, the inverse orthogonal transform unit is replaced by a function approximation unit
96
B. The function approximation unit
96
B performs function approximation using coefficients obtained by the parameter decoder
95
, thereby calculating approximate data of original data for each section.
In general, the orthogonal transform unit
93
A and inverse orthogonal transform unit
96
A are identical circuits as hardware, and are different only in calculation contents.
Also, the function approximation units
93
B and
96
B are identical circuits as hardware.
However, in each conventional data compression method, the value (sample value) of each sample point after sampling is quantized (digital) data, and the data is compressed by operating the digital value. Compression using this quantized digital data causes a rounding error at the least significant bit by a plurality of times of processing in compression. The compression algorithm decreases the bit precision of the sample value.
A higher compression ratio rapidly decreases the approximate degree of the waveform after compression/reconstruction with respect to the original waveform (sampling string). The quality of reconstructed data greatly degrades with an increase in compression ratio.
In addition, the conventional digital data compression must execute a large number of digital numerical calculations in function approximation, orthogonal transform, and encoding.
For example, letting k be the number of points in the section, 2k product-sum calculations or klog 2k product-sum calculations even using a high-speed calculation method must be done for orthogonal transform.
Similarly, product-sum calculation must be done for reconstruction. When the above high-speed calculation method is used, the number k of calculation points must satisfy 2J=k (J is a positive integer).
Even in processing of compressing data by substituting it into the algebraic curve in the real space, many calculations must be done to search for a coefficient almost free from any error in approximation calculation using the algebraic curve.
For this reason, when audio data and image data are to be compressed, a high-speed A/D converter, high-speed MPU, digital signal processor (DSP), and the like are required to shorten the data compression time in application fields necessary for very-high-speed processing, e.g., real-time compression.
Hence, an integrated circuit (IC) dedicated for each processing must be developed and manufactured, and a high-speed MPU must be used for processing. This increases the product development cost, product cost, and power consumption.
SUMMARY OF THE INVENTION
The present invention has been made to overcome the conventional drawbacks, and has as its object to provide a data compression method and apparatus capable of compressing an analog signal at high precision with a simple circuit arrangement without performing many calculations.
To achieve the above object, according to the present invention, there is provided a data compression method of sequentially sampling an analog input signal at discrete sample points, classifying a plurality of obtained sample values into a plurality of sections, and compressing the analog input signal, comprising the steps of setting a sampling string made up of sample values included in each section as a first sampling string of the section, generating a function which takes a sample value at a predetermined sample point included in the first sampling string and is approximated to the first sampling string, outputting a parameter defining the function as an element of compressed data of the section, and generating a second sampling string made up of sample values obtained at respective sample points by the function, performing predetermined calculation between two sample values at identical sample points in the first and second sampling strings for each section to calculate a new sample value, thereby creating for each section a third sampling string in which a new sample value at an arbitrary sample point is 0, and comparing for each section the third sampling string with a plurality of reference patterns set as reference sampling strings in advance, selecting a reference pattern most approximate to the third sampling string, and outputting a pattern code representing the selected reference pattern as an element of compressed data of the se
Allen Kenneth R.
NuCore Technology Inc.
Townsend and Townsend / and Crew LLP
Young Brian
LandOfFree
Data compression method and apparatus does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Data compression method and apparatus, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Data compression method and apparatus will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2547638