EP0898752A1

EP0898752A1 - Process and device for processing a signal

Info

Publication number: EP0898752A1
Application number: EP97923820A
Authority: EP
Inventors: Klaus Lehmann; Peter Kartmann; Christian Heilmann
Original assignee: IAD GmbH; Siemens AG; IAD Gesellschaft fuer Informatik Automatisierung und Datenverarbeitung mbH
Current assignee: IAD GESELLSCHAFT fur INFORMATIK AUTOMATISIERUNG
Priority date: 1996-05-15
Filing date: 1997-05-15
Publication date: 1999-03-03
Also published as: US6069975A; WO1997043721A1; DE19619572A1; CA2254879A1

Abstract

To process a signal (s(t)), particularly to recognize discontinuity, an orthogonal imaginary signal (s(t)) is first formed based on signal ((s)t). Then, a transformational step follows upon which a differentiation step follows. Processed in this way, the signal (s(t)) shows the discontinuity especially clearly, so that it can be selected and detected. Preferred applications of this process and the devices which operate according to it are telecommunications technology and signal processing in grids, especially in low, high, and medium tension technology.

Description

description

Method and device for processing a signal

The invention relates to a method and a device for processing a signal, in particular for detecting a discontinuity in the signal.

A discontinuity in a signal, in particular an analog measurement signal, is often difficult to detect or determine in practice, especially if the discontinuity is very small compared to the signal. As a rule, the signal m must be converted in a suitable manner for this purpose. In the present case, discontinuity is understood to mean, in particular, a disturbance in the signal, a superimposed fault signal or a discontinuity due to signal influencing or falsification. The signal itself can be present as analog or digital information.

A known approach to reshaping the signal to form a complex time signal is the Hilbert transformation, which is used, for example, in communications engineering to form a so-called analytical signal.

In this regard, reference is made, for example, to the article "Wigner distribution as a tool for time-frequency analysis of power station signals" from TM Technisches Messen 61 (1994) 1, pages 7 to 15. The method mentioned there provides for the linking of signals shifted to one another in the manner of autocorrelation and their transformation.

The invention is based on the object of specifying a method and a device for processing a signal, if a small discontinuity in the signal is to be recognized more easily than before using simple means.

According to the invention, the task with regard to the method is solved with the following steps: a) an orthogonal image signal is formed for the signal, b) the quotient between the image signal and the signal is formed, c) this quotient signal is stored on a finite set of values mapped by means of a mapping function, and d) the mapped quotient signal is differentiated.

An alternative method according to the invention provides the following steps: a) an orthogonal image signal is formed for the signal, b) a common mean signal z (t) is formed as a weighted function of time from the signal and the image signal, and c) the mean signal z ( t) is differentiated.

With simple means, these two methods allow an improved discernibility of a discontinuity in the signal such that it stands out disproportionately and is therefore easier to detect. The methods can be carried out very quickly with a computer and with only a small amount of computation.

The is preferably used to form the mean value signal z (t)

Relationship z (t) , where kl and k2 are weighting factors, s (t) the signal and s (t) the imaginary signal. The factors k1 and k2 can be specified by the person skilled in the art depending on the application and can be the same or different. This type of signal processing is very simple and can also be implemented with simple analog components if necessary. lize. In the case of a digital version, only little computing effort is required.

The method can be followed by a second and / or a third differentiation step. In this way, the discontinuity, depending on the application, can be highlighted even better from the signal course.

The signal which is differentiated em or multiple times can advantageously be subjected to pattern recognition, so that there is an accurate detection of the discontinuity, in particular a fault. Optionally, after or upon detection of a discontinuity, an alarm signal can be generated, which according to the prior art can be fed to further information processing or forwarding.

The methods are preferably used in medical technology. This applies in particular to the processing of an on-lme phonocardiogram, a late potential signal or an EEG signal.

The methods can also advantageously be used in communications technology, in particular for use in the demodulation of a frequency shift key signal or frequency shift keymg signal (FSK signal). Discontinuities can thus be recognized and dealt with quickly and reliably. Treatment can generally be understood to mean fading out, selecting or filtering out, in which case the discontinuity or the fault signal can optionally be selectively fed to further processing so that it serves as an information signal.

The methods are also advantageously used in measurement signal processing in electrical power supply, in particular in low, medium or high voltage technology, a current or voltage signal being used as the signal. 4 signal of a power supply network can serve. The short processing times and the safe working method are of particular advantage here. Special applications are meter technology (flow and electricity measurement) and network protection technology.

The further object with regard to the device is achieved according to the invention with a device for processing a signal, which may have a discontinuity, with a) a device for forming an orthogonal imaginary signal from the signal, b) a divider, which has a quotient signal between the Imaginary signal and the signal forms, c) a mapping element, which maps the quotient signal to a finite set of values by means of a mapping function, and d) a first differentiator, to which the quotient signal shown is fed.

An alternative device according to the invention for processing a signal has the following features: a) a device for forming an orthogonal imaginary signal from the signal, b) an averager which forms a common mean signal as a weighted function of time from the signal and the imaginary signal , and c) a first differentiator to which the mean signal is fed.

The advantages given above for the methods apply analogously to these devices. The first differentiating element can be followed by a second differentiating element and, if necessary, the third differentiating element can be followed by a third differentiating element. 5 be. The discernibility of a discontinuity is therefore particularly good.

A module for pattern recognition can be connected downstream of the differentiator, which module may also emit a message signal when a discontinuity is detected. The pattern recognition can be implemented, for example, with the aid of a filter.

The device is preferably designed with a computer, that is to say as a digital processing device, the device and the elements used for signal processing being designed as program modules. The advantages of the invention can thus be used particularly well.

An essential basic idea of the invention consists in that, from a mathematical point of view, apparently unnecessarily complexly shaped time signals also form quotients in the time domain and through number transformation from the number space

-oo <- <oo, e.g. the arctan operation leads to a finite, reasonably editable defined number space, for example —π <φ <π.

Contrary to the usual signal and system theory, in which the known functions of magnitude, phase and group delay are basically functions of angular frequency or frequency, the thermal bath or signals processed are functions of time.

The method is based on the formation of an imaginary part s (k * T _A ) for a sampled signal curve or signal s (k * T _Λ ) to be examined. For this purpose, the imaginary signal, s [k * T _Λ ) is determined or calculated orthogonally. It arises 6 time-dependent complex signal z (k * T _A ) = s {k * T _A ) + js {k * T _A ). In other words, this means that the newly calculated orthogonal imaginary signal is interpreted as an imaginary part.

The following can be used to calculate this new signal component or the Imagmar signal:

- the Hilbert transformation of s {k * T _A ) (known per se from communications engineering), - the differentiation of s [k * T _A ),

- the integration of s (k * T _A },

M-l

- the moving averaging of s (k * T _A ), ie ^ s [(k -v) * T _A J as υ = ü approximately offset-insensitive integration.

Herein, k em runtime index and T _A mean the time between two sampling points.

The interpretation of the complex time signal z \ k * T _A ) as

Magnitude z (k * T _A ) = js ² (k * T _A ) + s ² (k * T _A ) (geometric mean) and

Phase φ additionally allows the calculation the angular velocity φ (k * T _A ) by one time differentiation of φ [k * T _A ) and the angular acceleration φ (k * T _A } by two times differentiation of φ [k * T _A ).

The signal z (k * T _A ) formed can be subjected to further processing steps in order to obtain the so-called radius 7 slowness z \ k * T _A ) due to the unique differentiation of z (k * T _A ) and the radius acceleration z \ k * T _A ) due to two differentiations

Differentiation of z (k * T _A ) to form.

All sizes described and derived are functions of time.

Depending on the complexity of the task or the application of the respective signal processing, ie determining small discrepancies in a segment of a signal s {k * T _A ), the sizes described are sufficient to solve the present analysis task. In addition, for example, linked r—. - ds (k * T _A ) r -,

Sizes like yjz ² (k * T _A ) + φ ² (k * T _A ), * H {s (k * T _A ) \

0 {AI _A ).

Examples of options for designing and evaluating discontinuities in a technical signal curve include:

Detection of the attraction of an armature in the case of DC or AC voltage relays from the current profile,

Determination of the switching steps carried out by a stepper motor from the current profile,

Discrepancies in position and displacement measuring systems from the sensor signal, determination of the instantaneous frequency in FSK-modulated signals for fast data transmission, and

- Determination of irregularities in rotatory processes.

Further application possibilities can be found in medicine, for example in the case of irregularities in the activity of the heart valve in the so-called phonocardiogram and in the signal display and analysis of "late potentials". These are multi-dimensional ECG signals that indicate complications in patients after a heart attack.

Exemplary embodiments, further advantages and details of the invention are explained in more detail below with reference to the drawing. Show it:

FIG. 1 shows a block diagram of the device,

FIG. 2 shows a detail of the device according to FIG. 1 and FIGS. 3 to 12 waveforms for various application examples of the method and the device.

1 shows a block diagram of a device 1 for processing a signal s (t). A signal s (t) provided with one or more discontinuities or a periodic sequence sp (t) of signal sections is converted into a discrete number sequence by means of an analog-digital converter 5 in a filter 3 after suitable analog pre-filtering scan information T _{A is} supplied.

By means of an orthogonal signal transformation, a component or an imaginary signal becomes a real sampled signal s (k * T _A ) in the discrete area that as

"Imaginary part" of s (k * T _Λ ) can be viewed. This can be done, for example, in a separate transformer 7 or in a computer 9. The transformer 7 can also be formed by the computer 9, for example as a program module. The device 1 also has a memory 13 for storing programs, data and signals and a human-machine interface, for example a screen 11. W

9 formation by suitable further Signaltranεformationen as magnitude, differentiation, integration, arctangent formation and combinations of such operations, applied to the ge ^¬ formed complex signal z (k * T _A) = s [k * T _A) + js [k * T _A), lasεen εich small discontinuities in the course of the signal s (k T _A) ^¬ advantageous way to extract and determine *, in particular, when, whether and how often such features occurred.

Depending on the real-time requirement of the processing, the necessary signal transformations and links can be carried out by special hardware and software on universal computers.

2 shows a block diagram of the functions of a processing device, in particular of the computer 9, in detail. This has a device 14 for forming an orthogonal imaginary signal from the signal ε (t). This is followed by a processing module 15. This can be designed as an averager in the sense described above or as a divider with a downstream mapping element. The operations or transformations listed in the introduction can serve as mapping functions.

A differentiating module 16 is provided below, which can comprise one or more differentiating elements or differentiating functions. The differentiated output signals of the differentiating module 16 can then be fed to a module 17 for pattern recognition. This can, for example, be equipped with a filter function for the selection or detection of the discontinuity. The output signal can then be, for example, a warning signal w, a selected signal se or the input signal which has been cleaned of the discontinuity. In the following, more details on the new processes and the corresponding devices are explained in the specific exemplary embodiments.

example 1

3 shows a general sine signal with 4 periods of amplitude 1 and in the second half period of the sine signal a high-frequency sine signal with 5% of the amplitude is added. In the sum signal according to FIG. 4, this "discontinuity" due to the high-frequency sine signal can no longer be recognized.

Two examples of processing according to the present method clearly show when and for how long this "discontinuity" occurred.

darctan- ^s

In Fig. 5, φ = - at the top and in the lower track

_z - shown. In this signal processing, the dt desired information about the discontinuity is contained in the radius velocity z as well as in the speed of rotation φ of the complex signal z formed and is clearly emphasized.

Example 2

In a triangular signal according to FIG. 5, which can be, for example, an EEG signal from medical technology, small irregularities are added in the rising part by a one percent Smuss interference. 5 already shows the sum signal.

After processing with the new method, in particular the derivation, the discontinuities in the signal curves of φ in the upper track and in φ m lower are shown in FIG _

11 track apparently. These discontinuities, which are already optically recognizable, can then be detected and detected using simple detection means according to the prior art.

In FIG. 8, the product syk * T _A ) * H |, y (/ c * T _A ) | is in the upper lane as a linked size and plotted in the lower track z. 9 shows in the upper track z, in the lower track φ, in which case the imaginary part of the signal is generated not by silver transformation but by differentiation. All signal curves show a clear emergence of the fault.

Example 3

This example according to FIG. 10 applies, for example, to the demodulation of a 3-tone FSK signal (frequency shift keying signal), the relative frequency spacing being exactly as large as the data rate (here 4.8). H. there is a bandwidth efficiency

Bit enz of 1 before. \ s * Hz)

To determine the instantaneous frequency φ, we normalize to the square of the envelope z, ie the instantaneous frequency φ is formed according to the following rule: ds (k * T _A ) dH {s (k * T _A )} φ _N = H {s (k * T)} * s {k * T _A ) d (k * T _A ) d (k * T _A )

10 shows the FSK signal in the upper track and φ _N in the lower track. Here φ _{N is} directly proportional to the instantaneous frequency. As can easily be seen, the data rate could be increased further at the same frequency spacings. Further signal processing can be advantageous to improve noise sensitivity. The advantage according to the invention of this demodulation method is that in addition to modern frequency-selective methods according to the prior art, there is also a time selection criterion with which the so-called three-dimensional short-term spectrum can be determined more precisely and more reliably. Frequency-selective separation of the exemplary signal cannot be carried out at a data rate equal to the frequency spacing.

Example 4

In this example, the current profile of an AC voltage protection when switched on according to the proposal according to the invention was recorded as an analog signal and evaluated digitally. FIG. 11 shows the current curve, FIG. 12 shows the signal curve of φ normalized to the square of the envelope. It is easy to see that and when the anchor has tightened and that there is still a slight bouncing.

Examples of options for using the method and the device for a technical signal curve include:

- Determination of the executed switching steps of a stepper motor from the current curve,

Discontinuities in position and displacement measuring systems from the sensor signal,

- determination of saturation in a transducer,

Determination of a fault signal in network protection technology, determination of the instantaneous frequency in FSK-modulated signals for fast data transmission, and

- Determination of irregularities in rotatory processes. W

A special application is e.g. in medicine, e.g. in the event of irregularities in the activity of the heart valves, shown in the so-called phonocardiogram, and in the signal display and analysis of "late potentials". These are multi-dimensional ECG signals that indicate complications in patients after a heart attack.

Another application relates to measurement signal processing in electrical energy supply, in particular in low, medium or high voltage technology, it being possible for a current or voltage signal from an energy supply network to serve as the signal. The short processing times and the safe working method are of particular advantage here. Special applications are meter technology (flow and electricity measurement) and network protection technology, whereby a combination in connection with a communications technology application can also result.

Claims

14 claims

1. Method for processing a signal (s (t)), which may have a discontinuity, with the following steps: a) An orthogonal imaginary signal (s (t)) is formed for the signal, b) the quotient between the imaginary signal ( s (t)) and the signal (s (t)) are formed, c) this quotient signal is mapped onto a finite range of values by means of a mapping function, and d) the quotient signal shown is differentiated.

2. Method for processing a signal, which may have a discontinuity, with the following steps: a) an orthogonal imaginary signal {s (t)) is formed for the signal (s (t)), b) from the signal (s (t )) and the imaginary signal (.v (t)) forms a common mean signal z {t) as a weighted function of time, and c) the mean value signal z (t) is differentiated.

3. The method according to claim 2, wherein to form the mean signal z (t) the relationship z (t) serves, where Kl and K2 are weighting factors, s (t) the signal and .v (t) the imaginary signal.

4. The method according to claim 3, wherein the factors Kl and K2 are the same or different.

5. The method according to any one of the preceding claims, wherein a second differentiation step follows.

6. The method of claim 5, wherein a third connecting Diffe ^¬ renzierschritt.

7. The method according to any one of the preceding claims, wherein the signal differentiated one or more times is subjected to a pattern recognition.

8. The method according to claim 7, wherein a signal or warning signal (w) is generated when a discontinuity is detected.

9. The method according to any one of the preceding claims for use in medical technology.

10. A method for use in processing an on-line phonocardiogram, a late potential signal or an EEG signal.

11. The method according to any one of claims 1 to 10 for use in the demodulation of a Frequency Umtastεsignal (FSK signal).

12. The method according to any one of claims 1 to 11 for use in measuring signal processing in electrical power supply, in particular in low, medium or high voltage technology.

13. The method according to claim 12, wherein the signal used is a current or voltage signal from an energy supply network.

14. The method according to claim 13, wherein a differentiated signal is fed to a protective or counting device.

15. Device for processing a signal, which may have a discontinuity, with: 16 a) A device (7, 14) for forming an orthogonal

Imaginary signal (s (t)) from the signal (s (t)), b) a divider (15), which forms a quotient signal between the imaginary signal (.v (t)) and the signal (s (t)) , c) a mapping element (9, 15) which maps the quotient signal to a finite set of values by means of a mapping function, and d) a first differentiating element (16) to which the quotient signal shown is fed.

16. Device for processing a signal, which may have a discontinuity, with a) a device (14) for forming an orthogonal imaginary signal (s (t)) from the signal (s (t)), b) an averager (15), which forms a common mean value signal as a weighted function of time from the signal (s (t)) and the imaginary signal (.y (t)), and c) a first differentiator (16) to which the mean value signal is fed.

17. The apparatus of claim 15 or 16, wherein the first differentiating element (16) is followed by a second differentiating element or the first differentiating element (16) comprises a second or further differentiating function.

18. The apparatus of claim 17, wherein the second differentiating element is followed by a third differentiating element.

19. Device according to one of claims 15 to 18, wherein a differentiation member (16) is followed by a module (17) for pattern recognition.

20. Device according to one of claims 15 to 19, with a

Computer (9), the device (14) and the signal

Processing elements used (15,16,17) are executed as program modules.