JPH0548467A - Data compression/expansion system for time series signal - Google Patents

Abstract

PURPOSE:To compress a time series signal data outputted from a spectrum analyzer. CONSTITUTION:An FID being an output of an NMR is given to a Fourier transformation means 1, and a spectral location revision means 2 rearranges the location of the obtained spectrums in the order of their peak level from a low frequency to a high frequency in a descending order. Then an inverse Fourier transformation means 3 restores the data again into a time series data and a differentiation means 4 obtains a difference from a preceding value and data compression processing is applied to the difference signal by the known high efficiency coding system such as the Huffman coding.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a time series signal data compression / expansion system for compressing / expanding a time series signal output from a spectroscopic analyzer.

[0002]

2. Description of the Related Art As a spectroscopic analysis device, various devices such as a nuclear magnetic resonance device (NMR) and an infrared spectroscopic device have been put into practical use. And, in the past, the spectroscopic analysis has repeatedly repeated the operation of collecting and analyzing a certain amount of data, but in recent years, since high-speed processing can be performed by a computer, all the data must be collected. There is a tendency to analyze data. However, since the computer is not always available at the time of collecting data, the collected data is stored in a floppy disk (FD).
Or hard disk (HD) or magnetic tape (M
T) or the like is stored in the storage means. In addition, the collected data is also transferred to another workstation or the like via a network such as a local area network (LAN).

[0003]

However, when all the data is collected and then the data is analyzed as described above, the amount of collected data tends to be enormous. Management is becoming an issue.
In fact, taking NMR as an example, in recent years, multidimensional NMR has been put to practical use, for example, tens of megabytes for three-dimensional NMR and hundreds of megabytes for four-dimensional NMR. On the contrary,
Currently, since the capacity of HD is about 300 Mbytes, MT is usually used when storing a huge amount of data, but MT is inconvenient to handle and data is taken in for analysis. However, there is a problem that the analysis time becomes long because the time required for the data access, that is, the data access time is long. Further, the data transfer rate of the LAN is usually several tens of kilobytes / second to 1 megabyte / second, so that the cost is high when transferring the enormous amount of data as described above through the LAN.

By the way, it is generally known that the amount of data is reduced by performing data compression, which allows a large amount of data to be stored in a storage device such as an HD and the data transfer time. It can be shortened. Therefore, it is understood that there is a possibility that the above problem can be solved by performing data compression on the time series signal output from the spectroscopic analysis device. However, the time-series signal output from the spectroscopic analyzer is different from ordinary image data and audio signals, and in the end, important information is what frequency has what kind of peak and what size has a peak. However, the data compression method that has been conventionally used for compressing image data and audio signals cannot be used as it is. In the case of image data and audio signals, it is usual to allow some error when decompressing compressed data. However, regarding the data of the spectroscopic analysis, as described above, what frequency is used? Since there is a peak of a certain size or its state needs to be completely restored, an error is not allowed in data compression / expansion.

Further, it is desirable that the time required for data compression / decompression is short. That is, the total time of the data compression / expansion time and the data access time, that is, the sum of the time required for data compression, the time required for reading data from the storage device, and the time required for data expansion when storing data in the storage device. , When transferring data again,
It is desirable that the total of the time required for data compression, the time required for data transfer, and the time required for data decompression be shorter than before. If this is not the case, the efficiency of the entire spectroscopic analysis will decrease even if the data is compressed. Therefore, even if a data compression / decompression method with very few errors can be constructed by performing complicated calculations,
It is very difficult to adopt such a method, because the processing time for data compression / decompression becomes very long and the cost of the apparatus becomes very high.

The present invention is to solve the above problems, and has a simple structure, a short processing time for compression / expansion, and a data compression / expansion of a time-series signal with a small error when restored. The purpose is to provide a method.

[0007]

In order to achieve the above object, the data compression method of the time series signal of the present invention converts the time series signal into a signal in the frequency domain to increase the peak position from the low frequency side to the high frequency side. After the rearrangement is performed so as to be simply reduced to the frequency side, the time series signal is converted again, the difference between the time series signal and the previous value is obtained, and the compression processing is performed on the difference signal, The data expansion method of the time-series signal according to the present invention converts the time-series signal into a signal in the frequency domain and rearranges the peak position from the low frequency side to the high frequency side so that the peak position is simply reduced again. It is converted to a signal, the difference between the time series signal and the previous value is obtained, the data obtained by compressing the difference signal is decoded, then integrated and converted into a frequency domain signal, and the peak position is calculated as Repositioned to And converting the sequence signal.

[0008]

The time-series signal is converted into a signal in the frequency domain by the Fourier transforming means 1, and the spectrum position changing means 2 shifts the positions of the obtained spectra in order of peak height from the low frequency side to the high frequency side. The rearrangement is done so that it simply decreases to. After that, the inverse Fourier transform means 3 converts the time series data again, the differentiating means 4 obtains the difference from the previous value, and the difference signal is subjected to data compression processing.
The compressed data is decoded by the decoding processing means 6, then the difference signal is integrated by the integration means 7, and the Fourier transform means 8 transforms it into a frequency domain signal. After that, each peak is returned to its original position by the spectrum rearranging means 9 and converted into the original time series signal by the inverse Fourier transforming means 10.

[0009]

Embodiments will be described below with reference to the drawings.
In the following embodiments, an example of a time series signal (FID) output from NMR is taken as a time series signal, but the present invention is applied to time series signals output from other spectroscopic analysis devices such as infrared spectroscopic devices. Of course, it can be applied.

FIG. 1 shows the configuration of an embodiment in which the time-series signal data compression method according to the present invention is applied to a one-dimensional NMR time-series signal.
Is a spectral position changing means, 3 is an inverse Fourier transform means,
Reference numeral 4 is a differentiating means, and 5 is a compression encoding processing means.

The FID output from NMR (not shown) is first converted into a signal in the frequency domain by the Fourier transform means 1 to obtain the spectrum of the FID, and the spectrum position changing means 2 determines the position of the obtained spectrum. Rearrange from the low frequency side to the high frequency side in the order of peak height. Now, assuming that the FID spectrum obtained by the Fourier transforming means 1 is as shown in FIG. 2A, the spectrum position changing means 2 detects the peak frequency and its level, and detects the peak A having the maximum level as, for example, positive. Is placed in the vicinity of DC on the frequency side (center frequency is F ₁ ) and the peak B having the second level is placed in the vicinity of DC on the negative frequency side (center frequency is F ₂ ) and the third level The peak having the fourth level is located near the positive high frequency side of the peak A (center frequency is F ₃ ), and the peak having the fourth level is near the negative high frequency side of the peak B (center frequency is F ₄ ). All peaks are sequentially processed in the order of their peak levels. At this time, it is desirable to arrange the peaks as close as possible on the frequency axis. This is because the degree of data compression can be increased by this. As a result, as shown in FIG. 2B, the spectrum has an arrangement in which the positive and negative low frequencies are simply decreased from the low frequencies to the high frequencies with the direct current as the center. At this time, the spectrum position changing means 2
Creates information indicating the correspondence between the center frequencies before and after the spectrum rearrangement as shown in FIG. 3, for example. In the process of rearranging the spectrum, f _W in FIG.
The cut-out width of the peak indicated by may be a predetermined fixed value, or may be determined for each peak based on the half-value width of the peak, for example. In the latter case, the center frequency after rearrangement differs depending on the width of each peak, but in the former case, the center frequency after rearrangement can be fixed. It is also possible to divide the spectrum into blocks for each predetermined frequency width and rearrange the arrangement based on the order of the maximum level peaks in each block.

Although the FID spectrum has positive and negative frequencies, similarly, in the case of a spectrum having only positive frequencies, the peak envelope curve is changed from the low frequency side to the high frequency side as shown in FIG. The arrangement is changed so that the spectrum is simply reduced. Further, in FIG. 2A, P _{1 to} P ₆
As for the signal between the peaks shown by, since the level is low, it can be ignored, but the obtained FID
Is required to be accurately compressed / decompressed, it is arranged in the frequency region indicated by Q and Q ′ in FIG. 2B, that is, in the frequency side higher than the rearranged peaks. It is needless to say that the arrangement order of the low-level signals can be performed in the order of the levels, like the rearrangement of peaks described above.
Since the level of the signal existing between these peaks is low and almost the same,
For example, P ₃ , P ₂ , P ₁ on the negative high frequency side of peak F in FIG. 2B
Signals are arranged in order, and P ₄ ,
The signals of P ₅ and P ₆ may be simply arranged in the order of the spectrum, such as being arranged in order. Needless to say, in this case, it is desirable to arrange them as close as possible on the frequency axis. Obviously, the above processing can be performed by using a microcomputer or the like.

The data rearranged by the spectral position changing means 2 is converted back into time series data by the inverse Fourier transform means 3 and differentiated by the differentiating means 4. This differentiating process is performed by obtaining the previous value, that is, the difference from the immediately preceding sampling value. The differential signal obtained by the differentiating means 4 is subjected to data compression by the compression encoding processing means 5. For this data compression, a conventionally known high-efficiency coding method for a differential signal such as Huffman coding can be used. The compression / encoding processing means 5 outputs, together with the compressed data, information indicating the correspondence between the center frequencies before and after the rearrangement made by the spectrum position changing means 2 as shown in FIG. The output of the compression encoding processing means 5 is accumulated in a storage device or transferred via a network.

When compressing the differential signal, it is obvious that the peak of high signal strength is arranged on the low frequency side as described above. This is because, when a peak with high signal intensity exists on the high frequency side, the amplitude change of the difference signal from the previous value becomes large, and the compression rate becomes small even if the compression process is performed. Of course, it is possible to reduce the amplitude change of the differential signal by narrowing the sampling interval, but in this case, the data amount of the differential signal increases. On the other hand, when the peaks are rearranged so as to be simply reduced from the low frequency side to the high frequency side as described above, the low frequency component has a large signal strength, and the higher the frequency component, the higher the signal strength becomes. Because it gets smaller
The change in the amplitude of the differential signal is small, and therefore the compression rate can be improved without narrowing the sampling interval.

The data compressed as described above is shown in FIG.
The data is decompressed by the data decompression method having the configuration shown in.
In FIG. 5, 6 is a decoding processing means, 7 is an integrating means, 8 is a Fourier transforming means, 9 is a spectrum rearranging means, and 10 is an inverse Fourier transforming means. The data compressed by the data compression method shown in FIG. 1 is decoded by the decoding processing means 6,
It is integrated by the integrating means 7. This integration process is shown in FIG.
The inverse transform of the differentiation performed by the differentiating means 4 is performed by sequentially adding the decoded difference signal to the immediately preceding sampling value.

The output of the integrating means 7 is Fourier transformed by the Fourier transform output 8, and the spectrum rearrangement output 9 restores the spectrum arrangement to the original state. For example, assuming that the output of the Fourier transform output 8 is as shown in FIG. 2B, and the center frequencies of the peaks before and after the rearrangement of the spectrum have the correspondence relationship as shown in FIG. 3, the spectrum rearranging means 9 Based on the information in FIG. 3, the peaks A to F shown in FIG. 2B are rearranged to their original positions. This results in Figure 2A
The spectrum shown in is restored. Then, the restored spectrum is returned to the time-series signal by the inverse Fourier transform means 10 to restore the FID.

Although one embodiment of the present invention has been described above, the present invention is not limited to the above embodiment, and various modifications can be made. For example, in the above embodiment, one-dimensional N
The compression / expansion of the FID which is the output of the MR has been described, but it is possible to perform the compression / expansion of the data by performing the above processing on each of the plurality of time-series signals output from the multidimensional NMR. it is obvious. Further, instead of compressing the time-series signal at once, the time-series signal can be divided into several pieces and compressed. By performing such dynamic compression, it is possible to perform compression and disk access or compression and transfer via a network in parallel, and it is possible to reduce the total time required for data input / output. It's clear.

[0018]

As is apparent from the above description, according to the present invention, the data amount of the time series signal can be compressed to 1/2 to 1/3. In fact, 1 / for standard FID data
It has been confirmed that it can be compressed to about 2. Therefore, it is possible to reduce the capacity for accumulating data, and it is possible to accumulate more data in one storage device. Further, it is possible to shorten the read time of data from the storage device and the transfer time in the case of data transfer as compared with the conventional case. And further,
Even considering the time required for Fourier transform and its inverse transform, and the time required for spectrum rearrangement and rearrangement processing, the time required for data input / output can be significantly shortened compared to the past. is there.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of an embodiment of a time-series signal data compression method of the present invention.

FIG. 2 is a diagram for explaining spectrum rearrangement.

FIG. 3 is a diagram illustrating an example of information indicating a correspondence relationship between frequencies of respective peaks before and after spectrum rearrangement.

FIG. 4 is a diagram illustrating spectrum rearrangement for a time-series signal having only a positive frequency.

FIG. 5 is a diagram showing a configuration of an embodiment of a time-series signal data expansion system of the present invention.

[Explanation of symbols]

1 ... Fourier transforming means, 2 ... Spectral position changing means,
3 ... Inverse Fourier transforming means, 4 ... Differentiating means, 5 ... Compression encoding processing means, 6 ... Decoding processing means, 7 ... Integrating means, 8 ...
Fourier transforming means, 9 ... Spectrum rearranging means, 10 ...
Inverse Fourier transform means.

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁵ Identification code Internal reference number FI technical display location G01N 30/86 D 8105-2J

Claims (2)

[Claims] 1. A time series signal is converted into a frequency domain signal and rearranged so that the peak position is simply reduced from a low frequency side to a high frequency side, and then converted into a time series signal again, and the time series signal is converted. A data compression method for a time-series signal, wherein a difference from a previous value is obtained in a signal, and the difference signal is compressed. 2. A time-series signal is converted into a signal in the frequency domain, rearrangement is performed so that the peak position is simply reduced from the low frequency side to the high frequency side, and then the time-series signal is converted again, and the time series signal is converted. In the signal, the difference from the previous value is obtained, the data obtained by performing compression processing on the difference signal is decoded, then integrated and converted into a signal in the frequency domain, and the peak position is rearranged to the original position. A data decompression method for time-series signals, which is characterized by converting into a series signal.