PatentDe  


Dokumentenidentifikation EP1156446 21.12.2006
EP-Veröffentlichungsnummer 0001156446
Titel System und Verfahren zur Klassifikationseignungsbeurteilung von Merkmalen
Anmelder Siemens AG, 80333 München, DE
Erfinder Maetschke, Stefan, 90518 Unterrieden, DE
DE-Aktenzeichen 60124339
Vertragsstaaten AT, CH, DE, ES, FR, GB, LI
Sprache des Dokument EN
EP-Anmeldetag 15.05.2001
EP-Aktenzeichen 011117710
EP-Offenlegungsdatum 21.11.2001
EP date of grant 08.11.2006
Veröffentlichungstag im Patentblatt 21.12.2006
IPC-Hauptklasse G06K 9/62(2006.01)A, F, I, 20051017, B, H, EP

Beschreibung[en]
FIELD OF THE INVENTION

The present invention relates generally to the field of signal processing. More specifically, the present invention relates to signal processing a characteristic signal of a subject.

BACKGROUND OF THE INVENTION

In industrial automation, signal processing is used to classify an object being manufactured or processed based on a characteristic of the object. For example, an apple might be classified by a weight sensor configured to sense the weight of the apple. If the weight is greater than a predetermined weight, the apple is identified as "good", and, if not, the apple is identified as "bad".

However, the object can also be classified by other signals. For example, the apple might also be classified by acquiring a color digital image of the apple. If the apple is darker than a predetermined gray scale, or if the apple lacks sufficient red color, the apple is identified as "bad". The challenge is to determine which characteristic (e.g., weight, color, gray scale, etc.) best classifies the objects into the desired classifications, so that the best characteristic can be used during production to automatically classify objects.

US-4658429 discloses a system and method for preparing a recognition dictionary. For the preparation of a tree structure recognition dictionary, a feature set at each node of the tree should be the feature giving the largest discrete distribution number. Elements of an object are classified or divided into categories, whereas this classification is effected, when the distance between distributions for a certain feature is larger than a predetermined value.

A standard method for evaluating the classification of objects has been implemented which assumes a bimodal distribution of the measured characteristic, the distributions assumed to be Gaussian. For example, referring to FIG. 1, this standard method generates a histogram 10 of the frequency of occurrence of different values of the characteristic. The x-axis represents the values of the characteristic (e.g., weight, color, etc.) and the y-axis represents the frequency of objects having that characteristic. A first mode 12 includes objects in a first class (e.g., "bad" objects) and a second mode 14 includes objects in a second class (e.g., "good" objects). According to this method, the mean values 17, 18 of each mode are identified, the variances of mean values 17, 18 are determined, and the distance 19 between mean values 17 and 18 is determined. The smaller the variances and the greater the interval between mean values 17, 18, the greater is the quality of the characteristic for classification of this object.

One drawback of this method is that characteristic distributions frequently are neither bimodal nor Gaussian and, thus, are incorrectly evaluated by this prior method. With reference to FIG. 2, a frequency distribution 20 of another characteristic is shown, in which mode 22 is not Gaussian. Further, mode 22 includes objects in a first class, mode 24 includes objects in a second class, and mode 26 includes additional objects in the first class. An example of such a distribution might be one in which the characteristic is the length of a wooden dowel, wherein "good" dowels must have a length within a certain tolerance. Thus, "bad" dowels have lengths greater than (mode 26) and less than (mode 22) "good" dowels (mode 24). Prior methods will not adequately evaluate the suitability of this characteristic for classification purposes, since the distribution in FIG. 2 is not Gaussian and not bimodal.

Accordingly, there is a need for a system and method for evaluating the suitability of characteristics for classification. There is further a need for such a system and method which is applicable to non-Gaussian distributions. Further still, there is a need for such a system and method which is applicable to non-bimodal distributions. There is also a need for such a system and method which is robust against noise.

SUMMARY OF THE INVENTION

According to an exemplary embodiment, a method of evaluating a characteristic for suitability in classification of subjects based on subject data is provided. The subject data includes characteristic data and class data. The method includes arranging the subject data along an axis based on the values of the characteristic data, and identifying the number of class changes from one class to another class in the arranged subject data. The number of class changes represents the suitability of the characteristic for classification of the subjects.

According to an alternative embodiment, a method of evaluating a characteristic for suitability in classification of subjects based on subject data is provided. The subject data includes characteristic data and class data. The method includes arranging the subject data along an axis based on the values of the characteristic data, identifying consecutive subject data having a class change, and measuring the interval between the two consecutive subject data. The interval between class changes represents the suitability of the characteristic for classification of the subject.

According to yet another alternative embodiment, a system for evaluating a characteristic for suitability in classification of subjects is provided. The system includes sensing means for acquiring characteristic data from a plurality of subjects and classification means for classifying each subject with one of a first class and a second class. The system further includes means for arranging the subject data along an axis based on the values of the characteristic data and identifying the number of class changes from one class to another class in the arranged subject data. The number of class changes represents the suitability of the characteristic for classification of the subjects.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention will become more fully understood from the following detailed description, taken in conjunction with the accompanying drawings, wherein like reference numerals refer to like elements, and in which:

  • FIG. 1 is a histogram of a bimodal, Gaussian frequency distribution of a characteristic;
  • FIG. 2 is a histogram of a non-bimodal, non-Gaussian frequency distribution of a characteristic;
  • FIG. 3 is a block diagram of a system for evaluating a characteristic for suitability in classification of objects, according to an exemplary embodiment;
  • FIG. 4 is a flow diagram showing steps in a method according to an exemplary embodiment;
  • FIG. 5 is a number ray according to an exemplary embodiment;
  • FIG. 6 is a number ray illustrating a characteristic relatively good for classification; and
  • FIG. 7 is a number ray illustrating a characteristic relatively bad for classification.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

Referring first to FIG. 3, a block diagram of an exemplary system is shown. System 10 includes a plurality of sensors 12 and a signal processing circuit 14, and, optionally, a display 16 and user input device 18. System 10 acquires characteristic data sets (e.g., an image data set, a sound data set, other one-dimensional or multi-dimensional signal data sets, etc.) of a subject 20 (e.g., an object such as a welding spot, a tile, a motor, etc.) on a conveyor belt or platform 22. System 10 may be implemented on a multisignal processing system, such as, SIMULTAN, manufactured by Siemens AG, Munich, Germany. The SIMULTAN system can generate many different characteristic data sets regarding a subject and can process a wide variety of signal types such as, images, sounds, vibration, current, force, etc. Thus, the system and method described below selects the best characteristic or characteristics to use for each classification purpose.

Sensors 12 (e.g., video cameras, ultrasonic transducers, infrared sensors, microphones, etc.) are configured to acquire characteristic data, for example, in the form of a digital image of subject 20 against a background or platform 22. In this example, a video camera is configured to acquire an image data set in gray scale and to transmit the image data set to signal processing circuit 14. Alternatively, the video camera may acquire characteristic data in color and transmit separate subject data sets for red, green, and blue for each image. Alternatively, only one sensor may be available, though multiple sets of characteristic data may be generated or calculated from the output of the one sensor.

Signal processing circuit 14 (i.e., a computer) includes a programmed microprocessor (e.g., an INTEL x86 processor, microcontroller, etc.), memory, communication interfaces, etc. Alternatively, signal processing circuit 14 may comprise programmable logic, discrete circuit components, etc. Circuit 14 operates according to a software or hardware program stored in memory (e.g., hard disk storage, firmware, non-volatile memory, etc.), is configured to perform various signal processing functions on the received characteristic data sets, and may further be configured to provide display signals to display 16 (e.g., a cathode ray tube, liquid crystal display, etc.) and receive user input from user input device 18 (e.g., a keyboard, touchscreen, etc.).

Referring now to FIG. 4, steps performed by circuit 14 will be described. At step 102, system 10 is configured to sense, via sensors 12, characteristic data regarding subject 20. According to one example, a user places subject 20 (e.g., a metal part having a spot weld) on platform 22 and presses a switch (e.g., button, mouse click, touch screen, etc.) on user input device 18. In response, signal processing circuit 14 commands sensors 12 to sense one or more characteristics of subject 20. For example, one of sensors 12 might be a video camera configured to obtain a digital image of subject 20. Another of sensors 12 might be a weight sensor configured to weigh subject 20. Yet another of sensors 12 might be a length sensor configured to measure a diameter of subject 20. Any type of sensor for measuring any characteristic about a subject 20 is contemplated. When the one or more characteristic data sets of subject 20 are acquired, they are transmitted to signal processing circuit 14 which stores the characteristic data in a data structure in memory for further processing. The data structure includes both characteristic data and any other subject data, such as class data (see step 104 below). The user typically processes a number of subjects (e.g., 10, 20-30, etc.) to acquire a statistically significant sampling of the various classes of subjects.

At step 104, the subject data is classified into one of a number of classes. In this exemplary embodiment, the subject data is classified into a "good" class and a "bad" class. Alternatively, subject data can be classified into classes representing various sizes (e.g., short, medium, long), qualities, features, etc. One method of classifying subject data is for a user to press a switch on user input device 18 to indicate whether the subject currently on platform 22 is good or bad. A user puts a plurality of subjects on platform 22 and, for each subject, acquires characteristic data, and associates a class (e.g., good spot weld, bad spot weld, etc.) with each subject to generate the class data. Circuit 14 is configured to store the subject data, having the characteristic data and the class data, in a memory in preparation for further processing.

At step 106, circuit 14 is configured to begin processing of the subject data. Circuit 14 first arranges the subject data based on the values of the characteristic data. More specifically, the characteristic data typically assumes values over a range of values from a minimum to a maximum (e.g., from a deep red color to a deep blue color, from a minimum weight of zero to a maximum weight of perhaps 2-3 kilograms, etc.). As a graphical representation of the result of this arrangement, a number ray 30 is shown in FIG. 5. Number ray 30 includes a line 32 extending outward from a minimum characteristic value 34. Each of the subject data is plotted on number ray 30, such as is shown at points "X" 36 and "O" 38. The "X" and "O" indicia represent the class data for each data point (e.g., class X, class O), as generated at step 104. The number ray is merely a representation of the arrangement; typically, in an actual embodiment, circuit 14 is configured to store the subject data in a linear data structure or in consecutive memory locations to form the arrangement. Alternative arrangement methods are contemplated.

At step 108, circuit 14 is configured to identify the number of class changes from one class to another in the arranged subject data. Referring again to FIG. 5, circuit 14 is configured to read the class data from each subject data, proceeding from point 34 along line 32 on number ray 30. At point 40, a class "X" is read. The next consecutive or neighboring point along line 32, point 42, indicates a class "O", i.e., a change in class from class X to class 0. Circuit 14 is configured to identify this as a class change and to increment a class change counter (e.g., a memory location). Circuit 14 continues along number ray 30 until it reaches point 44, a class "O" point, and point 46, a class "X" point. Circuit 14 identifies this as another class change and again increments the class change counter. Circuit 14 continues in this manner until all or enough of subject data is read. The number of class changes represents the suitability of this characteristic for classification of subject 20. If a large number of class changes exist, then the characteristic is less suitable for classification of subject 20. If a small number of class changes exist, then the characteristic is more suitable for classification of subject 20.

At step 110, a further advantageous feature is shown. Circuit 14 is further configured to measure the interval between class changes in the arranged subject data, such as intervals 48 and 50 in FIG. 5. The greater the intervals between class changes, the more suitable the characteristic is for classification of subject 20. The lesser the intervals between class changes, the less suitable the characteristic is for classification of subject 20. As can be seen, either or both of the number of class changes and the size of the intervals may be used by circuit 14 to determine the suitability of the characteristic for classification of subject 20.

At step 112, circuit 14 is configured to generate a suitability value for each characteristic based on the number of class changes and/or the size of the intervals in the arranged subject data. The suitability value can then be used to compare the several characteristics of subject 20 relative to one another to determine which one is most suitable for classification. According to one exemplary equation, the suitability value is calculated as follows: i = 1 n 1 d i + a

wherein i = an index, n = the number of class changes, d i = the interval between class changes, and a is a constant used to vary the relative weight of the number of class changes versus the size of the intervals. This equation accommodates for the situation wherein d(i) = 0, i.e., two subject data have the same characteristic value but different classes. The smaller this suitability value, the better the characteristic is for classifying the subjects. The larger this suitability value, the worse the characteristic is for classifying the subjects.

At step 114, one or more of the characteristics is selected as most suitable for classification of subject 20. Circuit 14 may be configured to perform this step automatically by simply comparing the suitability values, or this may be done by a user via user input device 18 after viewing the suitability values, number of class changes, and/or intervals between class changes on display 16. Once steps 102-114 are complete, circuit 14 is configured to automatically classify subjects, for example, during production, using the best characteristic or group of characteristics for classification purposes.

FIGS. 6 and 7 are number rays 120 and 130 illustrating a first characteristic being relatively good for classification and a second characteristic 130 being relatively bad for classification, respectively. Number ray 120 includes only four class changes, each separated by at least a small interval. Number ray 130 includes approximately eleven class changes, most separated by a very small interval.

The measured characteristics of the subjects, as mentioned, can be any type of characteristic about the subjects which can be sensed or computed. In addition to those mentioned, circuit 14 may be configured to calculate further characteristics based on sensed characteristic data. For example, from a digital image of the subject, circuit 14 may calculate the mean value, variance, diameter, standard deviation, etc. of points within the digital image, each of which is a further characteristic of the subject which may be suitable for classification purposes. As another example, a microphone may receive a sound signal from the subject. One method of testing the quality of a tile is to tap the tile with a hammer and record the sound resonating therefrom. This sound signal is a characteristic of the subject, and data derived from the sound signal, such as, maximum amplitude, frequency, decay time, square root, absolute square, etc., are further characteristics of the subject which may be suitable for classification purposes. The system and method disclosed in FIGS. 3-5 above identifies which of these characteristics is best suited for classification of the tiles, for example, into "good" and "bad" parts.

While the exemplary embodiments have been illustrated and described, it should be understood that the embodiments disclosed herein are offered by way of example only. For example, the subject data, comprising characteristic data and class data, may be stored and arranged in various types of data structures and/or on various types of memories. Further, in addition to two-classification schemes (e.g., "good" parts and "bad" parts), subjects may be classified into three, four, or more classifications. The invention is not limited to a particular embodiment, but extends to various modifications that nevertheless fall within the scope of the appended claims.


Anspruch[de]
Auf einem Computer ausgeführtes Verfahren zur Beurteilung eines Merkmals in bezug auf Eignung bei der Klassifikation von Subjekten auf der Basis gemessener physikalischer Subjektdaten, wobei die Subjektdaten Merkmaldaten und Klassendaten umfassen, mit den folgenden Schritten: Anordnung der gemessenen physikalischen Subjektdaten auf einer Achse auf der Basis der Werte der Merkmaldaten; und Identifizieren der Anzahl von Klassenänderungen von einer Klasse zu einer anderen Klasse in den angeordneten gemessenen physikalischen Subjektdaten, wobei die Anzahl der Klassenänderungen die Eignung des Merkmals für die Klassifikation der Subjekte repräsentiert. Verfahren nach Anspruch 1, ferner mit dem Schritt des Messens des Intervalls zwischen Klassenänderungen in den angeordneten gemessenen physikalischen Subjektdaten, wobei das Intervall zwischen Klassenänderungen die Eignung des Merkmals für die Klassifikation der Subjekte repräsentiert. Verfahren nach Anspruch 2, ferner mit dem Schritt des Erzeugens eines Geeignetheitswerts gleich i = 1 n 1 d i + a

wobei i = ein Index, n = die Anzahl der Klassenänderungen, di = das Intervall zwischen Klassenänderungen und a ist eine Konstante.
Verfahren nach Anspruch 1, wobei die Merkmaldaten Tondaten umfassen. Verfahren nach Anspruch 1, ferner mit dem Schritt des Klassifizierens der gemessenen physikalischen Subjektdaten in die erste Klasse oder die zweite Klasse. Verfahren nach Anspruch 1, wobei die gemessenen physikalischen Subjektdaten zweite Merkmaldaten und zweite Klassendaten umfassen, ferner mit den folgenden Schritten: Anordnung der gemessenen physikalischen Subjektdaten auf der Basis der zweiten Merkmaldaten, um zweite angeordnete gemessene physikalische Subjektdaten zu erzeugen; und Identifizieren der Anzahl von Klassenänderungen von einer Klasse zu einer anderen Klasse in den zweiten angeordneten gemessenen physikalischen Subjektdaten, wobei die Anzahl der Klassenänderungen in den zweiten angeordneten gemessenen physikalischen Subjektdaten die Eignung des zweiten Merkmals für die Klassifikation der Subjekte repräsentiert. Verfahren nach Anspruch 6, ferner mit dem Schritt des Messens des Intervalls zwischen Klassenänderungen in den zweiten angeordneten gemessenen physikalischen Subjektdaten, wobei das Intervall zwischen Klassenänderungen in den zweiten angeordneten gemessenen physikalischen Subjektdaten die Eignung des zweiten Merkmals für die Klassifikation der Subjekte repräsentiert. Verfahren nach Anspruch 6, ferner mit dem Schritt des Auswählens des ersten oder des zweiten Merkmals, das weniger Klassenänderungen aufweist, als das Merkmal, das die Subjekte am geeignetsten repräsentiert. Verfahren nach Anspruch 1, ferner mit den folgenden Schritten: Identifizieren aufeinanderfolgender gemessener physikalischer Subjektdaten mit einer Klassenänderung; und Messen des Intervalls zwischen den zwei aufeinanderfolgenden gemessenen physikalischen Subjektdaten, wobei das Intervall zwischen Klassenänderungen die Eignung des Merkmals für die Klassifikation des Subjekts repräsentiert. Verfahren nach Anspruch 9, wobei die Merkmaldaten einen Durchmesser des Subjekts umfassen. Verfahren nach Anspruch 9, ferner mit dem Schritt des Klassifizierens der gemessenen physikalischen Subjektdaten in die erste Klasse oder die zweite Klasse. Verfahren nach Anspruch 9, wobei die Merkmaldaten Bilddaten umfassen. Verfahren nach Anspruch 9, wobei die gemessenen physikalischen Subjektdaten zweite Merkmaldaten und zweite Klassendaten umfassen, ferner mit den folgenden Schritten: Anordnung der gemessenen physikalischen Subjektdaten auf der Basis der zweiten Merkmaldaten, um zweite angeordnete gemessene physikalische Subjektdaten zu erzeugen; und Identifizieren der Anzahl von Klassenänderungen von einer Klasse zu einer anderen Klasse in den zweiten angeordneten gemessenen physikalischen Subjektdaten, wobei die Anzahl der Klassenänderungen in den zweiten angeordneten gemessenen physikalischen Subjektdaten die Eignung des zweiten Merkmals für die Klassifikation der Subjekte repräsentiert. Verfahren nach Anspruch 13, ferner mit dem Schritt des Messens des Intervalls zwischen Klassenänderungen in den zweiten angeordneten gemessenen physikalischen Subjektdaten, wobei das Intervall zwischen Klassenänderungen die Eignung des zweiten Merkmals für die Klassifikation des Subjekts repräsentiert. Verfahren nach Anspruch 14, ferner mit dem Schritt des Auswählens des ersten oder des zweiten Merkmals, das die Subjekte am geeignetsten klassifiziert, auf der Basis der Anzahl von Klassenänderungen und des Intervalls zwischen Klassenänderungen für jedes des ersten und des zweiten Merkmals. Verfahren nach Anspruch 9, wobei die gemessenen physikalischen Subjektdaten auf einer Achse angeordnet werden, wobei die aufeinanderfolgenden Subjektdaten benachbarte Positionen auf der Achse aufweisen. System zur Beurteilung eines Merkmals in bezug auf Eignung bei der Klassifikation von Subjekten, umfassend: Meßmittel zum Erfassen von Merkmaldaten von einer Vielzahl von Subjekten; Klassifikationsmittel zum Klassifizieren jedes Subjekts mit einer ersten Klasse oder einer zweiten Klasse; und Mittel zum Anordnen der gemessenen physikalischen Subjektdaten auf einer Achse auf der Basis der Werte der Merkmaldaten und zum Identifizieren der Anzahl von Klassenänderungen von einer Klasse zu einer anderen Klasse in den angeordneten gemessenen physikalischen Subjektdaten, wobei die Anzahl der Klassenänderungen die Eignung des Merkmals für die Klassifikation für die Subjekte repräsentiert. System nach Anspruch 17, wobei die Mittel zum Anordnen und Identifizieren eine Signalverarbeitungsschaltung umfassen. System nach Anspruch 17, ferner mit Mitteln zum Messen des Intervalls zwischen Klassenänderungen in den angeordneten gemessenen physikalischen Subjektdaten, wobei das Intervall zwischen Klassenänderungen die Eignung des Merkmals für die Klassifikation der Subjekte repräsentiert. System nach Anspruch 17, ferner mit Mitteln zum Erzeugen eines Geeignetheitswerts gleich i = 1 n 1 d i + a

wobei i = ein Index, n = die Anzahl der Klassenänderungen, di = das Intervall zwischen Klassenänderungen und a ist eine Konstante.
System nach Anspruch 17, ferner umfassend: Meßmittel zum Beschaffen zweiter Merkmaldaten von der Vielzahl von Subjekten; und Mittel zum Anordnen der gemessenen physikalischen Subjektdaten auf der Basis der zweiten Merkmaldaten, um zweite angeordnete gemessene physikalische Subjektdaten zu erzeugen, und zum Identifizieren der Anzahl von Klassenänderungen von einer Klasse zu einer anderen Klasse in den zweiten angeordneten gemessenen physikalischen Subjektdaten, wobei die Anzahl der Klassenänderungen in den zweiten angeordneten gemessenen physikalischen Subjektdaten die Eignung des zweiten Merkmals für die Klassifikation der Subjekte repräsentiert. System nach Anspruch 21, ferner mit Mitteln zum Messen des Intervalls zwischen Klassenänderungen in den zweiten angeordneten Subjektdaten, wobei das Intervall zwischen Klassenänderungen die Eignung des zweiten Merkmals für die Klassifikation der Subjekte repräsentiert. System nach Anspruch 22, ferner mit Mitteln zum Auswählen des ersten oder des zweiten Merkmals, das die Subjekte am geeignetsten klassifiziert.
Anspruch[en]
A method carried out on a computer for evaluating a characteristic for suitability in classification of subjects based on measured physical subject data, the subject data including characteristic data and class data, comprising: arranging the measured physical subject data along an axis based on the values of the characteristic data; and identifying the number of class changes from one class to another class in the arranged measured physical subject data, whereby the number of class changes represents the suitability of the characteristic for classification of the subjects. The method of claim 1, further comprising measuring the interval between class changes in the arranged measured physical subject data, whereby the interval between class changes represents the suitability of the characteristic for classification of the subjects. The method of claim 2, further comprising generating a suitability value equal to i = 1 n 1 d i + a

wherein i = an index, n = the number of class changes, d i = the interval between class changes, and a is a constant.
The method of claim 1, wherein the characteristic data includes sound data. The method of claim 1, further comprising classifying the measured physical subject data into one of the first class and the second class. The method of claim 1, the measured physical subject data including second characteristic data and second class data, further comprising: arranging the measured physical subject data based on the second characteristic data to create second arranged measured physical subject data; and identifying the number of class changes from one class to another class in the second arranged measured physical subject data, whereby the number of class changes in the second arranged measured physical subject data represents the suitability of the second characteristic for classification of the subjects. The method of claim 6, further comprising measuring the interval between class changes in the second arranged measured physical subject data, whereby the interval between class changes in the second arranged measured physical subject data, whereby the interval between class changes in the second arranged measured physical subject data represents the suitability of the second characteristic for classification of the subject. The method of claim 6, further comprising selecting the one of the first and second characteristic having fewer class changes as the characteristic which most suitably classifies the subjects. The method of claim 1, further comprising: identifying consecutive measured physical subject data having a class change; and measuring the interval between the two consecutive measured physical subject data, whereby the interval between class changes represents the suitability of the characteristic for classification of the subject. The method of claim 9, wherein the characteristic data includes a diameter of the subject. The method of claim 9, further comprising classifying the measured physical subject data into one of the first class and the second class. The method of claim 9, wherein the characteristic data includes image data. The method of claim 9, the measured physical subject data including second characteristic data and second class data, further comprising: arranging the measured physical subject data based on the second characteristic data to create second arranged measured physical subject data; and identifying the number of class changes from one class to another class in the second arranged measured physical subject data, whereby the number of class changes in the second arranged measured physical subject data represents the suitability of the second characteristic for classification of the subject. The method of claim 13, further comprising measuring the interval between class changes in the second arranged measured physical subject data, whereby the interval between class changes represents the suitability of the second characteristic for classification of the subject. The method of claim 14, further comprising selecting the one of the first and second characteristics which most suitably classifies the subjects based on the number of class changes and the interval between class changes for each of the first and second characteristics. The method of claim 9, wherein the measured physical subject data is arranged along an axis, the consecutive subject data having neighboring positions on the axis. A system for evaluating a characteristic for suitability in classification of subjects, comprising: sensing means for acquiring characteristic data from a plurality of subjects; classification means for classifying each subject with one of a first class and a second class; and means for arranging the measured physical subject data along on axis based on the values of the characteristic data and identifying the number of class changes from one class to another class in the arranged measured physical subject data, whereby the number of class changes represents the suitability of the characteristic for classification for the subjects. The system of claim 17, wherein the means for arranging and identifying includes a signal processing circuit. The system of claim 17, further comprising means for measuring the interval between class changes in the arranged measured physical subject data, whereby the interval between class changes represents the suitability of the characteristic for classification of the subjects. The system of claim 17, further comprising means for generating a suitability value equal to: i = 1 n 1 d i + a

wherein i = an index, n = the number of class changes, d i = the interval between class changes, and a is a constant.
The system of claim 17, further comprising: sensing means for acquiring second characteristic data from the plurality of subjects; and means for arranging the measured physical subject data based on the second characteristic data to create second arranged measured physical subject data and for identifying the number of class changes from one class to another class in the second arranged measured physical subject data, whereby the number of class changes in the second arranged measured physical subject data represents the suitability of the second characteristic for classification of the subjects. The system of claim 21, further comprising means for measuring the interval between class changes in the second arranged subject data, whereby the interval between class changes represents the suitability of the second characteristic for classification of the subject. The system of claim 22, further comprising means for selecting the one of the first and second characteristics which most suitably classifies the subjects.
Anspruch[fr]
Procédé mis en oeuvre sur un ordinateur pour évaluer une caractéristique quant à sa pertinence dans la classification de sujets, sur la base de données physiques mesurées de sujets, les données de sujets comprenant des données de caractéristiques et des données de classes, comprenant les étapes consistant à : disposer les données physiques mesurées de sujets, le long d'un axe, sur la base des valeurs des données de caractéristique ; et identifier le nombre de changements de classe d'une classe à une autre, parmi les données physiques mesurées de sujets disposées, d'où il résulte que le nombre de changements de classe représente la pertinence de la caractéristique pour la classification des sujets. Procédé suivant la revendication 1, comprenant en outre, le fait de mesurer l'intervalle entre des changements de classe parmi les données physiques mesurées de sujets disposées, d'où il résulte que l'intervalle entre des changements de classe représente la pertinence de la caractéristique pour la classification des sujets. Procédé suivant la revendication 2, comprenant en outre, le fait d'engendrer une valeur de pertinence égale à : i = 1 n 1 d i + a où i = un indice, n = le nombre de changements de classe, di = l'intervalle entre des changements de classe, et a est une constante. Procédé suivant la revendication 1, dans lequel les données de caractéristiques comprennent des données sonores. Procédé suivant la revendication 1, comprenant en outre, le fait de ranger les données physiques mesurées de sujets soit dans la première classe soit dans la deuxième classe. Procédé suivant la revendication 1, les données physiques mesurées de sujets comprenant des données de deuxième caractéristique et des données de deuxième classe, comprenant en outre, les étapes consistant à : disposer les données physiques mesurées de sujets, sur la base des données de deuxième caractéristique pour créer des deuxièmes données physiques mesurées de sujets disposées ; et identifier le nombre de changements de classe d'une classe à une autre, parmi les deuxièmes données physiques mesurées de sujets disposées, d'où il résulte que le nombre de changements de classe parmi les deuxièmes données physiques mesurées de sujets disposées représente la pertinence de la deuxième caractéristique pour la classification des sujets. Procédé suivant la revendication 6, comprenant en outre, le fait de mesurer l'intervalle entre des changements de classe parmi les deuxièmes données physiques mesurées de sujets disposées, d'où il résulte que l'intervalle entre des changements de classe parmi les deuxièmes données physiques mesurées de sujets disposées représente la pertinence de la deuxième caractéristique pour la classification du sujet. Procédé suivant la revendication 6, comprenant en outre, le fait de sélectionner celle de la première caractéristique et de la deuxième caractéristique comportant le moins de changements de classe, en tant que caractéristique permettant de classer les sujets de la manière la plus appropriée. Procédé suivant la revendication 1 comprenant en outre, les étapes consistant à : identifier des données physiques mesurées de sujets consécutives comportant un changement de classe ; et mesurer l'intervalle entre les deux données physiques mesurées de sujets consécutives, d'où il résulte que l'intervalle entre des changements de classe représente la pertinence de la caractéristique pour la classification du sujet. Procédé suivant la revendication 9, dans lequel les données de caractéristiques comprennent un diamètre du sujet. Procédé suivant la revendication 9, comprenant en outre, le fait de ranger les données physiques mesurées de sujets soit dans la première classe soit dans la deuxième classe. Procédé suivant la revendication 9, dans lequel les données de caractéristiques comprennent des données d'image. Procédé suivant la revendication 9, les données physiques mesurées de sujets comprenant des données de deuxième caractéristique et des données de deuxième classe, comprenant en outre, les étapes consistant à : disposer les données physiques mesurées de sujets, sur la base des données de deuxième caractéristique pour créer des deuxièmes données physiques mesurées de sujets disposées ; et identifier le nombre de changements de classe d'une classe à une autre, parmi les deuxièmes données physiques mesurées de sujets disposées, d'où il résulte que le nombre de changements de classe parmi les deuxièmes données physiques mesurées de sujets disposées représente la pertinence de la deuxième caractéristique pour la classification du sujet. Procédé suivant la revendication 13, comprenant en outre, le fait de mesurer l'intervalle entre des changements de classe parmi les deuxièmes données physiques mesurées de sujets disposées, d'où il résulte que l'intervalle entre des changements de classe représente la pertinence de la deuxième caractéristique pour la classification du sujet. Procédé suivant la revendication 14, comprenant en outre, le fait de sélectionner celle de la première caractéristique et de la deuxième caractéristique qui permet de classer de la manière la plus appropriée les sujets, sur la base du nombre de changements de classe et de l'intervalle entre des changements de classe pour chacune des première et deuxième caractéristiques. Procédé suivant la revendication 9, dans lequel les données physiques mesurées de sujets sont disposées le long d'un axe, les données de sujets consécutives ayant des positions voisines sur l'axe. Système pour évaluer une caractéristique quant à sa pertinence dans la classification de sujets, comprenant : des moyens de détection destinés à acquérir des données de caractéristiques à partir d'une pluralité de sujets ; des moyens de classification destinés à classer chaque sujet suivant une première classe ou une deuxième classe ; et des moyens destinés à disposer les données physiques mesurées de sujets le long d'un axe, sur la base des valeurs des données de caractéristique et à identifier le nombre de changements de classe d'une classe à une autre, parmi les données physiques mesurées de sujets disposées, d'où il résulte que le nombre de changements de classe représente la pertinence de la caractéristique pour la classification des sujets. Système suivant la revendication 17, dans lequel les moyens de disposition et d'identification comprennent un circuit de traitement de signaux. Système suivant la revendication 17, comprenant en outre, des moyens destinés à mesurer l'intervalle entre des changements de classe parmi les données physiques mesurées de sujets disposées, d'où il résulte que l'intervalle entre des changements de classe représente la pertinence de la caractéristique pour la classification des sujets. Système suivant la revendication 17, comprenant en outre, des moyens destinés à engendrer une valeur de pertinence égale à : i = 1 n 1 d i + a où i = un indice, n = le nombre de changements de classe, di = l'intervalle entre des changements de classe, et a est une constante. Système suivant la revendication 17 comprenant, en outre : des moyens de détection destinés à acquérir des données de deuxième caractéristique, à partir de la pluralité de sujets ; et des moyens destinés à disposer les données physiques mesurées de sujets, sur la base des données de deuxième caractéristique, pour créer des deuxièmes données physiques mesurées de sujets disposées et pour identifier le nombre de changements de classe, d'une classe à une autre, parmi les deuxièmes données physiques mesurées de sujets disposées, d'où il résulte que le nombre de changements de classe parmi les deuxièmes données physiques mesurées de sujets disposées représente la pertinence de la deuxième caractéristique pour la classification des sujets. Système suivant la revendication 21, comprenant en outre, des moyens destinés à mesurer l'intervalle entre des changements de classe parmi les deuxièmes données de sujets disposées, d'où il résulte que l'intervalle entre des changements de classe représente la pertinence de la deuxième caractéristique pour la classification du sujet. Système suivant la revendication 22, comprenant en outre, des moyens destinés à sélectionner celle de la première caractéristique et de la deuxième caractéristique qui permet de classer les sujets de la manière la plus appropriée.






IPC
A Täglicher Lebensbedarf
B Arbeitsverfahren; Transportieren
C Chemie; Hüttenwesen
D Textilien; Papier
E Bauwesen; Erdbohren; Bergbau
F Maschinenbau; Beleuchtung; Heizung; Waffen; Sprengen
G Physik
H Elektrotechnik

Anmelder
Datum

Patentrecherche

Patent Zeichnungen (PDF)

Copyright © 2008 Patent-De Alle Rechte vorbehalten. eMail: info@patent-de.com