EP1134712A1 - Method for the processing of the signal in a danger detector, and detector with means for the implementation of such method - Google Patents

Abstract

The signals from an alarm, comprising at least one sensor (2, 3, 4), for monitoring characteristic hazard values and an analytical electronic unit (1), connected to the at least one sensor (2, 3, 4), are compared with pre-set parameters. Furthermore, the signals are analysed for repeated or regular occurrence and repeated, or regularly occurring alarm signals are classed as error signals. The classification of signals as error signals gives rise to a corresponding adjustment of the parameter. When an error signal arises, before the parameter is adjusted, the validity of the signal analysis for the at least one sensor (2, 3, 4) is checked and the parameter adjustment is carried out, depending upon the result of said validity check. An alarm with the means for carrying out said method comprises at least one sensor (2, 3, 4), for a characteristic hazard value and an electronic analysis unit (1), containing a microprocessor (6), for the evaluation and analysis of the signal from the at least one sensor (2, 3, 4). The microprocessor (6) has a software programme with an adaptive algorithm based on multiple solutions for the analysis of the signals from the at least one sensor (2, 3, 4).

Description

The present invention relates to a method for processing the signals of a hazard detector, which has at least one sensor for monitoring hazard parameters and has evaluation electronics assigned to the at least one sensor, with the monitoring the hazard parameters by comparing the signals of the at least one sensor with predetermined parameters. The hazard detector can, for example, be a Smoke detectors, a flame detector, a passive infrared detector, a microwave detector, a dual detector (Passive infrared + microwave sensor), or a noise detector.

Today's hazard detectors have one with regard to the detection of hazard parameters Sensitivity achieved that the main problem is no longer a hazard parameter to detect as early as possible, but rather in the form of interference signals from real danger signals differentiate safely and thereby avoid false alarms. The distinction between danger and interference signals is essentially through the use several different sensors and correlation of their signals or by analysis different characteristics of the signals of a single sensor and / or by a corresponding one Signal processing, most recently through the use of fuzzy logic a significant improvement in interference immunity has been achieved.

Fuzzy logic is well known. With regard to the evaluation of signals from hazard detectors It should be emphasized that signal values known as fuzzy sets, or unsharp quantities, be assigned according to a membership function, whereby the value of the membership function, or the degree of belonging to a fuzzy set, between zero and one is. It is important that the membership function can be normalized, i.e. the sum all values of the membership function is equal to one, which results in the fuzzy logic evaluation allows a clear interpretation of the signal.

The present invention is now intended to be a method of the type mentioned at the outset for processing the signals of a hazard detector are given, which are insensitive to interference and immunity to interference is further improved.

The method according to the invention is characterized in that the signals of the at least a sensor is analyzed to determine whether they occur more frequently or regularly and that increasingly or regularly occurring signals are classified as interference signals.

A first preferred development of the method according to the invention is characterized in that that the classification of signals as interference signals a corresponding adaptation of the Parameter triggers.

The method according to the invention is based on the new finding that, for example, a Fire alarms between two revisions or two power failures never more than a few real fires "sees", and that increased or regular signals indicate the presence from sources of interference. The interference signals caused by the interference sources are recognized as such and the detector parameters are adjusted accordingly. In this way the detectors operated according to the method of the invention are capable of learning differentiate better between real danger signals and interference signals.

A second preferred development of the method according to the invention is characterized in that that if interference signals occur before the parameters are adjusted, the result the analysis of the signals of the at least one sensor is checked for its validity, and that the parameters are adjusted depending on the result of this validity check.

A third preferred development is characterized in that the validity check using methods based on multiple resolution.

A fourth preferred development of the method according to the invention is characterized in that that for the validation check Wavelets, preferably "biorthogonal" or "second generation "wavelets or" lifting scheme "can be used.

The wavelet transform is a transformation or mapping of a signal from the time domain in the frequency domain (see, for example, "The Fast Wavelet Transform" by Mac A. Cody in Dr. Dobb's Journal, April 1992); it is basically the Fourier transform and Fast Fourier transform similar. However, it differs from these by the basic function of the transformation after which the signal is developed. With a Fourier transform a sine and cosine function is used in the frequency domain is sharply localized and undetermined in the time domain. With a wavelet transformation a so-called wavelet or wave packet used. There are different types like this for example a Gaussian, Spline or Haar wavelet, each with two parameters can be shifted in the time domain and stretched or compressed in the frequency domain. Recently, new wavelet methods have been introduced, often as "second generation" be designated. Such wavelets can be used with the so-called "lifting scheme" (Sweldens) constructed.

A series of approximations of the original signal results, each a coarser one Resolution than the previous one. The number of operations required for the transformation is proportional to the length of the original signal, while at the Fourier transform this number is disproportionate to the signal length. The fast Wavelet transformation can also be done inversely by the original signal from the approximated values and coefficients for the reconstruction. The algorithm for the decomposition and reconstruction of the signal and a table of the coefficients disassembly and reconstruction are exemplified for a spline wavelet in "An Introduction to Wavelets "by Charles K. Chui (Academic Press, San Diego, 1992). See also "A Wavelet Tour of Signal Processing" by S. Mallat (Academic Press, 1998).

Another preferred development of the method according to the invention is characterized in that that the expected values for the approximation or approximation and Detailed coefficients of the wavelets are determined and compared at different resolutions become. The coefficients mentioned are preferably determined in an estimator or by means of a neural network.

The invention further relates to a hazard detector with means for performing the above Method, with at least one sensor for a hazard parameter and with one Microprocessor-containing evaluation electronics for evaluating and analyzing the signals of the at least one sensor.

The hazard detector according to the invention is characterized in that the microprocessor a software program with a multi-resolution learning algorithm for the Analysis of the signals of the at least one sensor contains.

This is a first preferred embodiment of the hazard detector according to the invention characterized in that on the one hand an analysis of the sensor signals mentioned by the learning algorithm on their repeated or regular occurrence and on the other hand a validity check the result of the analysis is done and that the learning algorithm for the validation Wavelets, preferably "biorthogonal" or "second generation" wavelets, are used.

This is a second preferred embodiment of the hazard detector according to the invention characterized that the learning algorithm uses neuro-fuzzy methods.

enthält, in denen ϕ_m,n Wavelet Skalierfunktionen, c_m,n Approximations-Koeffizienten und y_k den k-ten Eingangspunkt des neuronalen Netzes bezeichnet, und

die duale Funktion (dual function, Definition siehe S. Mallat) von ϕ_m,n ist.A third preferred embodiment of the hazard detector according to the invention is characterized in that the learning algorithm has the two equations f m (x) = Σ c m, n · Φ m, n (x) (Σ over all n) and

contains, in which ϕ _{m, n} wavelet scaling functions, c _{m, n} approximation coefficients and y _k denotes the k-th entry point of the neural network, and

is the dual function (definition see S. Mallat) of ϕ _{m, n} .

Fig. 1

ein Diagramm zur Funktionserläuterung,

Fig. 2

ein Blockschema eines mit Mitteln zur Durchführung des erfindungsgemässen Verfahrens ausgerüsteten Gefahrenmelders,

Fig. 3a, 3b

zwei Varianten eines Details des Gefahrenmelders von Fig. 2; und

Fig. 4

eine weitere Variante eines Details des Gefahrenmelders von Fig. 3.

The invention is explained in more detail below with the aid of exemplary embodiments and the drawings; it shows:

Fig. 1

a diagram to explain the function,

Fig. 2

2 shows a block diagram of a hazard detector equipped with means for carrying out the method according to the invention,

3a, 3b

two variants of a detail of the hazard detector of Fig. 2; and

Fig. 4

another variant of a detail of the hazard detector of FIG. 3.

The signals of a hazard detector are processed by the method according to the invention in such a way that that typical interference signals are recorded and characterized. If in the present Description primarily of fire detectors, it does not mean that the inventive Procedure is limited to fire detectors. The procedure is rather for hazard detectors suitable for all types, especially for intrusion and motion detectors.

The noise signals mentioned are analyzed using a simple and reliable method. An important feature of this method is that the interference signals are not only recorded and be characterized, but that the result of the analysis is checked. This will Wavelet theory and multiresolution analysis are used. Each according to the result of the check, the parameters of the detector or the algorithms customized. This means that, for example, the sensitivity is reduced or that certain automatic switching between different parameter sets can be locked.

The latter is explained using an example: In European patent application 99 122 975.8 a fire detector is described, which is an optical sensor for scattered light, a temperature sensor and has a fire gas sensor. The evaluation electronics of the detector contains one Fuzzy controller, in which a link between the signals of the individual sensors and a The respective type of fire is diagnosed. There is a special application-specific for each type of fire Algorithm provided and selectable based on the diagnosis. In addition, the Detector different parameter sets for personal protection and real estate protection, between which are normally switched online. If now with the temperature and / or If the fire gas sensor detects interference signals, the switchover between them Parameter sets locked.

When using fuzzy logic, one of the problems to be solved is translation of the knowledge stored in a database in linguistically interpretable fuzzy rules. Neurofuzzy methods developed for this purpose were unable to convince because they sometimes provide fuzzy rules that are very difficult to interpret. A way to In contrast, obtaining interpretable fuzzy rules offers so-called multiple-resolution techniques. Their idea is to use a dictionary of membership functions which form a multiple resolution, and determine which one for the description membership functions suitable for a control surface.

1 shows a diagram of such a multiple resolution. Line a shows the course a signal whose amplitude is in the small, medium and large ranges. Correspondingly, in line b, the membership functions c1 are "fairly small", c2 "medium" and c3 marked "rather large". These membership functions form a multiple resolution, which means that every membership function is a sum of membership functions a higher resolution level can be broken down. This results in those entered in line c Membership functions c5 "very small", c6 "small to very small", c7 "very medium", c8 "large to very large "and c9" very large ". According to line d, for example, the triangular one Spline function c2 in the sum of the translated triangular functions of the higher level of Row c can be disassembled.

In the Tagaki-Sugeno model, the fuzzy rules are based on the equation R i : if x is A i then y i = f i (x i ) expressed. Here, A _{i are} linguistic expressions, x is the input linguistic variable and y is the output variable. The value of the linguistic input variables can be sharp or fuzzy. For example, if x _{i is} a linguistic variable for temperature, the value x and can be a sharp number such as "30 (° C)" or a fuzzy quantity such as "about 25 (° C)", where "about 25" itself is a fuzzy set.

For a sharp input value, the output value of the fuzzy system is given by: y = Σβ i · F ( x ) / Σβ i where the degree of fulfillment β _{i is given} by the expression β _i = µ _Ai (x and), in which µ _Ai (x and) denotes the membership function for the linguistic term A _i . In many applications a linear function is used: f (x and) = a ^T i · x and + b _i . if a constant b _i is used to describe the sharp output value y, then the system becomes: R i : if x is A i then y i = b i

If spline functions N ^{k are} taken, for example as membership function µ _Ai (x and) = N ^k [2 ^m (x and-n)], then the system of equation (3) is equivalent to y i = Σb i · N k [2nd m ( x -n)]

In this particular case, the output y is a linear sum of translated and expanded Spline functions. And that means that under equation (4) the Tagaki-Sugeno Model is equivalent to a multi-resolution spline model. And it follows that here Wavelet techniques can be applied.

2 shows a block diagram of a danger detector equipped with a neurofuzzy learning algorithm. The detector designated by the reference symbol M is, for example, a fire detector and has three sensors 2 to 4 for fire parameters. For example, an optical one Sensor 2 for scattered light or transmitted light measurement, a temperature sensor 3 and a fire gas, for example, a CO sensor 4 is provided. The output signals of sensors 2 to 4 are fed to a processing stage 1 which contains suitable means for processing the signals, such as, for example, amplifiers, and pass from this into a subsequently as µP 6 designated microprocessor or microcontroller.

In µP 6, the sensor signals are interchangeable as well as individually with certain parameter sets compared for the individual fire parameters. Of course, the number the sensors are not limited to three. In this way, only a single sensor can be provided, in this case different properties from the signal of one sensor, for example the signal gradient or the signal fluctuation are extracted and examined. In the

In terms of software, µP 6 are a neuro-fuzzy network 7 and a validation check (validation) 8 integrated. If the signal resulting from the neuro-fuzzy network 7 is evaluated as an alarm signal an alarm delivery device 9 or an alarm center becomes a corresponding one Alarm signal supplied. If the validation 8 shows that repeated or regular Interference signals occur, then the parameter sets stored in µP 6 become corresponding corrected.

The neuro-fuzzy network 7 is a series of neural networks which use the symmetrical scaling functions ϕ _{m, n} (x) = ϕ _{m, n} (x) = ϕ [(xn) 2 ^m ] as an activation function. The scaling functions are such that {(ϕ _{m, n} (x)} form a multiple resolution. Each neural network uses activation functions of a given resolution. The mth neural network optimizes the coefficients c and _{m, n} with f _m (x ), the output of the mth neural network. f m (x) = Σ c m, n ϕ m, n (x) (Σ over all n)

wobei y_k(x) der k-te Eingangspunkt und

die duale Funktion von ϕ_m,n(x) ist. Die beiden Gleichungen (5) und (6) bilden den Hauptalgorithmus des Neuro-Fuzzy-Netzes.The coefficients c _{m, n} are calculated using the following equation:

where y _k (x) is the kth entry point and

is the dual function of ϕ _{m, n} (x). The two equations (5) and (6) form the main algorithm of the neuro-fuzzy network.

At each iteration step, the values of the various neural networks are cross-checked (validated), which is a property of the wavelet decomposition, namely that the approximation coefficients c _{m, n of} a level m from the approximation and wavelet coefficients of the level m-1 can be obtained using the reconstruction or decomposition algorithm.

eine Spline-Funktion zweiter Ordnung und ϕ_m,n(x) eine Interpolationsfunktion. Bei einer zweiten Ausführung ist ϕ_m,n(x) eine Spline-Funktion und

die duale Funktion von ϕ_m,n(x). In einer dritten Ausführung ist

= ϕ_m,n(x), wobei ϕ_m,n(x) die Haar-Funktion ist. In diesen Fällen ist die Implementierung des Lernalgorithmus in einen einfachen Mikroprozessor möglich.In a preferred embodiment

a second order spline function and ϕ _{m, n} (x) an interpolation function. In a second embodiment, ϕ _{m, n} (x) is a spline function and

the dual function of ϕ _{m, n} (x). In a third execution

= ϕ _{m, n} (x), where ϕ _{m, n} (x) is the hair function. In these cases, the implementation of the learning algorithm in a simple microprocessor is possible.

FIGS. 3a and 3b show two variants of a neuro-fuzzy network 7 and the associated validation stage 8. In the example of FIG. 3a, the input signal is approximated in different resolution levels as a weighted sum of wavelets ψ _{m, n} and scaling functions ϕ _{m, n} with a given resolution. The validation level 8 compares the approximation coefficients c and _{m, n} with the approximation and detail coefficients of the wavelets at the level of the next lower resolution level. P and q denote wavelet reconstruction filter coefficients.

In the example of FIG. 3b, the input signal is approximated in different resolution levels as a weighted sum of scaling functions ϕ _{m, n} with a given resolution. The validation level 8 compares the approximation coefficients c and _{m, n} with the approximation coefficients at the level of the next lower resolution level. With g, wavelet low-pass decomposition coefficients are designated.

basierende Dual-Spline-Estimatoren verwendet. Man verwendet Wavelet-Spline-Estimatoren zur adaptiven Bestimmung der geeigneten Auflösung, um eine zugrundeliegende Hyperfläche in einem online-Lernprozess lokal zu beschreiben. Ein bekannter Schätzer ist der Nadaraya-Watson-Estimator, mit welchem die Gleichung der Hyperfläche f(x) durch den folgenden Ausdruck abgeschätzt wird:

Instead of in a neuro-fuzzy network 7, the above-mentioned coefficients can be determined in an estimator of the type shown in FIG. 4. This estimator is a so-called multiple-resolution spline estimator, which is used to estimate the coefficients c and _{m, n} in the equation f _m (x) = c and _{m, n} ϕ _{m, n} (x) on the functions

based dual spline estimators. Wavelet spline estimators are used to adaptively determine the appropriate resolution in order to locally describe an underlying hypersurface in an online learning process. A well-known estimator is the Nadaraya-Watson estimator, with which the equation of the hypersurface f (x) is estimated by the following expression:

Nadaraya-Watson estimators have two interesting properties, they are estimators of the local mean square deviation and it can be shown that in the case of a random design they are so-called Bayesian estimators of (x _k , y _k ), where (x _k , y _k ) are iid copies of a continuous random variable (X, Y).

können als Schätzer verwendet werden. Wir verwenden zuerst die Funktion

zur Abschätzung von f(x) mit λ = 2^-m (m ist eine ganze Zahl) an x_n mit x_n·2^m ∈ Z: The spline functions ϕ (x) and their dual function

can be used as estimators. We use the function first

to estimate f (x) with λ = 2 ^-m (m is an integer) at x _n with x _n · 2 ^m ∈ Z:

ist Gleichung (6) für die duale Spline-Funktion äqivalent zur Verwendung eines bei x_n zentrierten Schätzers:

Using the symmetry of

Equation (6) for the dual spline function is equivalent to using an estimator centered at x _n :

The expected value of the counter in equation (7) is proportional to the approximation coefficient c _{m, n} . Equation (6) gives an estimate of c and _{m, n} in f _m (x) = Σc and _{m, n} ϕ _{m, n} (x): c m, n = f (x n ).

In Figure 4, the available data (values) are with a small square referred to, their projection onto dual spline functions with a small circle and the estimate on a regular grid with a cross.

wobei die Filterkoeffizienten g dem Tiefpass-Zerlegungs-Koeffizienten für Spline-Funktionen entsprechen. Ausserdem wird gefordert, dass

damit Teilungen durch sehr kleine Werte verhindert werden.Two conditions are necessary to validate the coefficient c and _{m, n} :

where the filter coefficients g correspond to the low-pass decomposition coefficient for spline functions. It is also required that

so that divisions are prevented by very small values.

The strength of this method is that the calculation of a coefficient c and _{m, n} requires only two values to be stored, the numerator and the denominator in equation (7). The method is therefore very well suited for online learning with a simple microprocessor with a small memory capacity.

The method is easily adaptable to density estimation by replacing equations (7) and (8) with the following equation:

Claims (11)

Method for processing the signals of a hazard detector, which has at least one sensor (2, 3, 4) for monitoring hazard parameters and evaluation electronics (1) assigned to the at least one sensor (2, 3, 4), in which a comparison of the signals of the at least one sensor (2, 3, 4) takes place with predetermined parameters, characterized in that the signals of the at least one sensor (2, 3, 4) are analyzed to determine whether they occur more or more regularly, and that more or more frequently occurring signals are classified as interference signals. A method according to claim 1, characterized in that the classification of signals as interference signals triggers a corresponding adjustment of the parameters. Method according to claim 2, characterized in that when interference signals occur before the adjustment of the parameters, the result of the analysis of the signals of the at least one sensor (2, 3, 4) is checked for its validity, and that the adjustment of the parameters as a function of The result of this validity check is carried out. Method according to claim 3, characterized in that the validity check is carried out by means of methods which are based on multiple resolution. Method according to claim 4, characterized in that wavelets, preferably "biorthogonal" or "second generation" wavelets or "lifting scheme" are used for the validity check. A method according to claim 5, characterized in that the expected values for the approximation or the approximation and detail coefficients of the wavelets are determined and compared at different resolutions. Method according to claim 6, characterized in that the said coefficients are determined in an estimator or by means of a neural network. Hazard detector with means for carrying out the method according to claim 1, with at least one sensor (2, 3, 4) for a hazard parameter and with evaluation electronics (1) containing a microprocessor (6) for evaluating and analyzing the signals of the at least one sensor (2, 3, 4), characterized in that the microprocessor (6) contains a software program with a learning algorithm based on multiple resolution for the analysis of the signals of the at least one sensor (2, 3, 4). Hazard detector according to claim 9, characterized in that the learning algorithm on the one hand analyzes the said sensor signals for their repeated or regular occurrence and on the other hand checks the validity of the result of the analysis, and that the learning algorithm for the validity check Wavelets, preferably "biorthogonal" or "second generation "wavelets. Hazard detector according to claim 9, characterized in that the learning algorithm uses neuro-fuzzy methods. Hazard detector according to claim 10, characterized in that the learning algorithm the two equations f m (x) = Σ c m, n · Φ m, n (x) (Σ over all n) and

contains, in which ϕ _{m, n} scaling functions, c and _{m, n} approximation coefficients and y _k denotes the k-th entry point of the neural network and

is the dual function of ϕ _{m, n} .

Images