WO1999014686A2 - Verfahren zur erfassung zeitabhängiger moden dynamischer systeme - Google Patents
Verfahren zur erfassung zeitabhängiger moden dynamischer systeme Download PDFInfo
- Publication number
- WO1999014686A2 WO1999014686A2 PCT/EP1998/005793 EP9805793W WO9914686A2 WO 1999014686 A2 WO1999014686 A2 WO 1999014686A2 EP 9805793 W EP9805793 W EP 9805793W WO 9914686 A2 WO9914686 A2 WO 9914686A2
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- modes
- prediction
- segmentation
- time series
- models
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 45
- 230000036962 time dependent Effects 0.000 title claims description 5
- 230000011218 segmentation Effects 0.000 claims abstract description 66
- 230000006870 function Effects 0.000 claims description 26
- 238000012549 training Methods 0.000 claims description 14
- 230000007958 sleep Effects 0.000 claims description 9
- 230000009467 reduction Effects 0.000 claims description 7
- 230000002829 reductive effect Effects 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 2
- 239000011521 glass Substances 0.000 claims 1
- 230000001537 neural effect Effects 0.000 abstract description 4
- 230000007704 transition Effects 0.000 description 34
- 230000008859 change Effects 0.000 description 8
- 230000008569 process Effects 0.000 description 7
- 238000004458 analytical method Methods 0.000 description 6
- 230000008901 benefit Effects 0.000 description 5
- 230000006978 adaptation Effects 0.000 description 4
- 230000006399 behavior Effects 0.000 description 4
- 238000001514 detection method Methods 0.000 description 4
- 239000000203 mixture Substances 0.000 description 4
- 230000000739 chaotic effect Effects 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 210000000601 blood cell Anatomy 0.000 description 2
- 230000009134 cell regulation Effects 0.000 description 2
- 230000002860 competitive effect Effects 0.000 description 2
- 238000011217 control strategy Methods 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 239000003814 drug Substances 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 238000012552 review Methods 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 230000002618 waking effect Effects 0.000 description 2
- 230000031018 biological processes and functions Effects 0.000 description 1
- 230000033228 biological regulation Effects 0.000 description 1
- 210000004369 blood Anatomy 0.000 description 1
- 239000008280 blood Substances 0.000 description 1
- 230000036772 blood pressure Effects 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000001311 chemical methods and process Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 210000003743 erythrocyte Anatomy 0.000 description 1
- 230000004424 eye movement Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000010327 methods by industry Methods 0.000 description 1
- 210000000056 organ Anatomy 0.000 description 1
- 230000029058 respiratory gaseous exchange Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000005204 segregation Methods 0.000 description 1
- 230000011664 signaling Effects 0.000 description 1
- 208000019116 sleep disease Diseases 0.000 description 1
- 230000002269 spontaneous effect Effects 0.000 description 1
- 230000005654 stationary process Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/0265—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion
- G05B13/027—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion using neural networks only
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/369—Electroencephalography [EEG]
- A61B5/372—Analysis of electroencephalograms
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/048—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators using a predictor
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B17/00—Systems involving the use of models or simulators of said systems
- G05B17/02—Systems involving the use of models or simulators of said systems electric
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/369—Electroencephalography [EEG]
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
- G10L25/12—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being prediction coefficients
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
- G10L25/30—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique using neural networks
Definitions
- the invention relates to a method for detecting dynamic systems that can be characterized by non-stationary system parameters over time, in particular a method for segmenting time series of measured variables (variables) of dynamic systems and for identifying the system parameters (modes) that characterize the segments.
- a dynamic system is understood here to mean in particular every occurrence, the time course of which by a discrete type
- ⁇ (t) denotes a set of characteristic system parameters
- x a state that generally forms a vector in a multidimensional state space
- y a state that is shifted in time.
- the state space is spanned by variables that e.g. B. can be physical, chemical, biological, medical, geological, geometric, numerical and / or process parameters.
- Parameters ⁇ can also be variable over time.
- a given system with parameters ⁇ that cannot be changed over time is also referred to below as mode.
- Observable or measurable system variables form detectable time series or data streams which are characteristic of the respective sequence of system modes. If the system parameters cannot be changed within the time series over certain time periods (segments), the time series can be subdivided according to the existing system modes (segmentation) and each segment can be assigned to a system mode (identification).
- An example of a system under consideration is the generation of speech signals in the mouth and throat area, in which the system constantly changes its configuration and thus its mode.
- a variable in the example: air pressure fluctuations
- dynamic systems can be analyzed on the basis of the measured signals, and a number of methods are known for obtaining models from time series which are suitable for predicting and controlling the system behavior. It is known, for example, that the state of a dynamic system can be modeled by recording the time dependence of observable measurands.
- this modeling is carried out by reconstructing the state space using so-called time-delay coordinates, as described, for example, in FIG. B. by NH Packard et al. in "Physical Review Letters” (Vol. 45, 1980, pp. 712 ff.).
- time-delay coordinates as described, for example, in FIG. B. by NH Packard et al. in "Physical Review Letters” (Vol. 45, 1980, pp. 712 ff.).
- the global reconstruction of the system is also disadvantageous because, in the case of applications for multidimensional systems, a large number of input variables have to be known in advance as boundary conditions and / or because of the high dimensionality, the system can practically no longer be estimated (recorded, mapped) and / or excessively high, impractical computing effort arises.
- the object of the invention is to improve methods for detecting the modes of dynamic systems with non-stationary Specify system parameters with which the limitations of conventional methods can be overcome and which, in particular, make it possible to automatically segment and identify time series with an increased number of details with practical processing effort and high reliability.
- the invention is based on the idea of considering transitions between different modes of a dynamic system as intermediate modes of the system, which represent linear interpolations of the output and end modes of the transition in pairs.
- the dynamic systems under consideration switch from one mode to the other rather gradually, instead of switching abruptly between modes.
- the invention aims to identify such transitions between dynamic modes in signals and the modes.
- a drift segmentation is carried out in which m each time segment, m the system passes from a first system mode s x to a second system mode s.
- a sequence of mixed Prediction models g x is detected, which is given by a linear, pair-wise superposition of the prediction models f of the two system modes s ⁇ .
- the invention also relates to a device for detecting a dynamic system with a large number of
- the device contains a device for recording a time series of at least one of the system variables x (t) des Systems, a switching segmentation device, which is set up to detect a predetermined prediction model f for a corresponding system mode s 1 in each time segment of a predetermined minimum length for the system variables x (t), and a drift segmentation device, with which in each time segment in which the system passes from a first system mode s x to a second system mode s, a sequence of mixed prediction models g x is detected.
- the device according to the invention can furthermore contain devices for setting interpolation and segmentation parameters, comparator circuits for processing prediction errors of prediction models, display and signaling devices and storage devices.
- the device according to the invention can be a monitor for physiological data or physical or chemical process parameters.
- Fig. 4 Curve representations for the segmentation of EEG data according to the inventive method.
- non-stationary time series are recorded using a two-step procedure in which a suitable modeling is carried out first and then a so-called drift segregation.
- the modeling is set up to record a predetermined prediction model for a corresponding system mode in each time segment of a predetermined minimum length for each system parameter.
- a conventional one is preferably used Switching segmentation, such as that from the publication by K. Pawelzik et al. m "Neural Computation" (Vol. 8, 1996, p. 340 ff.) is known.
- the modeling can, however, also be carried out by another procedure which is equivalent to the system information obtained for switching segmentation and which is adapted to a specific application, e.g. B. is adapted to known pure modes or boundary conditions.
- the switching segmentation is used to determine characteristic predictors that are suitable for describing the system modes.
- the switching segmentation can either be carried out on a tramings time series or on the time series to be examined. In both cases, the prediction models or predictors determined can be used for further, unknown time series.
- a dynamic system with a finite number N of different modes is considered. For the -th fashion is one
- the switching segmentation is found that the Time series ⁇ x t ⁇ divided according to the changing system modes.
- the functions f are derived as predictors (or: prediction models, expert functions) from a set of networks with variable parameters by means of a suitable training program in which both the parameters of the networks and the segmentation are determined simultaneously.
- the term "network” is used here for all possible suitable model functions, so preferably for neural networks, but also e.g. B. for polynomials or linear functional approximations.
- the optimal choice of a neural network depends on the specific application. Networks with a fast learning ability, such as e.g. B. so-called RBF networks (Radial Basis Function Network) of the Moody-Darken type are used.
- the training takes place on the condition that the system modes do not change with every time step, but have a lower switching rate, so that a system mode is retained over several time steps.
- the assumed limit of the switching rate or number of time steps over which a system mode is retained is initially a free input parameter and can be selected in a suitable manner depending on the application, for example depending on predetermined empirical values or on the basis of a parameter adaptation strategy.
- Training is done by maximizing the probability W that the set of networks would generate the time series ⁇ x t ⁇ . It is a training with competitive learning, as described in detail in the paper "Introduction to the theory of neural computation” by J. Hertz et al. (Addison-Wesley Publishing Company 1991, esp. Chap. 9: "Unsupervised competitive learning”).
- the application-dependent implementation of such training can be derived from this paper.
- the training rule of competing learning on the basis of the error occurring during learning can be represented according to (1).
- This training rule ensures that the learning speed (improvement of the parameters) is highest for the functions f with the smallest distance to the target value y.
- f x (x) 4x (lx) for xe [0, 1]
- f 3 (x) 2x for xe [0, 0.5) or
- f 3 (x) 2 (1-x) for xe [0.5, 1]
- f 4 (x) f 3 (f 3 ( ⁇ )
- the training results in a specialization of four of the predictors (6, 2, 4, 3) each in one of the above four modes.
- the stationary ranges are at the intervals [0, 50] and [400, 450] (fJ, [100, 150] (f 2 ), [200, 250] (f 3 ) and [300, 350] (f 4 )
- the other two predictors (3, 5) specialize in the transition areas between the modes, which shows the disadvantage of conventional switching segmentation, in which, in the case of transitions, the corresponding time range is divided several times without an adequate description.
- the transitions (so-called drifting, non-abrupt transition, sliding change) between the system modes are taken into account.
- drifting non-abrupt transition, sliding change
- the drift between system modes is thus as follows by an overlay of (or in pairs linear interpolation between) models exactly two modes. Mixed, possibly graded, intermediate modes occur, but these are not separate (pure) system modes.
- a set of P pure system modes, each represented by a network k (s), se P, and a set of M mixed system modes, each represented by a linear overlay of two networks i (s) and j ( s), se M, are represented.
- x is the vector (x t , x t . ⁇ , ..., x t . (M _ 1) ⁇ ) of the time delay coordinates of the time series ⁇ x t ⁇ and f i
- m is an embedding dimension and ⁇ the delay parameter of the embedding.
- the embedding dimension is the dimension of the phase space in which the system is viewed and in which the models operate.
- the resolution R corresponds to the number of permitted intermediate modes and is also used as a resolution or gradation of the interpolation designated between pure fashions.
- the resolution R can have any value, but is chosen to be sufficiently low, depending on the application, in order to achieve an optimal system description (especially in the case of very noisy processes) and practical computing times, in particular taking into account the switching rate mentioned above.
- the resolution R is selected manually by an operator or automatically by a control circuit as a function of an existing analysis result and a comparison with a predetermined threshold value.
- the total number of mixed modes for a given resolution R is between two networks
- R 'N • (Nl) / 2.
- 896.
- 896.
- the drift segmentation now includes the search for a segmentation with the pure and mixed system modes (a, b, R), which is optimized with regard to the prediction error of the modes of the entire time series.
- the predictors are selected so that each element of the time series can be assigned one of the modes from the total number of system modes.
- the prediction error is the deviation of a predictor prediction from the actual element of the time series to be examined.
- a prediction is determined for each time step with each of the predictors, which results in a time-dependent matrix of the predictor predictions, from which an average prediction error for arbitrarily selected segmentations can be derived.
- the segmentations The drift segmentation is the one with the smallest prediction error.
- the search for the segmentation with the least prediction error can be carried out using any suitable search or iteration technique.
- a dynamic programming technique is preferably selected which is equivalent to the Viterbi algorithm for HM models (so-called Hidden Markov Models). Details on this can be found, for example, in the publication "A tutorial on Hidden Markov Models and Selected Applications in Speech Recognition” by LR Rabiner in “Readings in Speech Recognition” (ed. A. Waibel et al., San Mateo, Morgan Kaufmann, 1990, p . 267-296).
- the drift segmentation is the most likely mode sequence within the HM models that the time series to be examined could have generated. As a secondary condition, the possibility of changing the mode is limited using the T function (see below).
- the aim of the adaptation is to specify an optimal sequence of networks or linear mixtures of these.
- a sequence is optimal if the so-called energy or cost function C * of the prediction is minimized.
- the cost function C * is composed of the sum of the quadratic errors of the prediction and the cost functions of the mode transitions of the sequence.
- the derivation of the cost function C * between two times t 0 and t max takes place inductively by first starting from a start cost function according to equation (4).
- T (s, s) is the cost function of the transition from a mode s to a mode s.
- the drift segmentation results from the determined optimal sequence of networks or linear mixtures of these, in that the modes that result in C * are traced back and recorded as a function of time.
- segmented modes are identified by assigning the associated system mode to each predictor or prediction model. This identification depends on the application.
- the drift segmentation comprises the search for a course a (t) which provides a special path between the pure modes for which the prediction error of the entire time series is optimized.
- FIG. 2 shows the occupation of the respective modes according to the networks determined as a function of time (time steps [1200, 2400]).
- the transition or drift areas are in accordance with their time limits and starting or. End modes are shown in frames in which the respective drift course between the modes is dotted.
- B in the time steps 1350 to 1400 between the networks 2 and 4.
- the transitions are linear as expected according to equation (12).
- R 3
- the segmentation shown in FIG. 2b results.
- the dotted transitions deviate from the linear drift course. Nevertheless, the representation is an adequate description of the dynamic behavior of the system, as the comparison of the temporal position of the modes and the third shows.
- Blood cell regulation in the human body represents a high-dimensional chaotic system which can be described by the Mackey-Glass-delay-differential equation (13) (see also in the above-mentioned treatise by J. Hertz et al.).
- time series of physiological parameters that are characteristic of the amount of red blood cells can be segmented depending on the application.
- the functionality of the segmentation is explained below as an example.
- Delay parameter ⁇ 1 (see equation (2)) gives the picture shown in FIG. 3b. The expected segmentation of the time series into stationary modes and drift transitions is shown.
- the reduction step comprises a sequential reduction in the number of networks, each associated with a determination of the average prediction error.
- the reduction (deduction of redundancy networks) is ended if a further reduction in the number of networks means a significant increase in the prediction error.
- Figure 3c shows the result of such a reduction.
- the mean square error RMSE remains constant when reduced by one, two, three and four networks, whereas there is a sharp increase in the case that only one network is used for modeling. This means that the system is optimally modeled with a number of networks that is equal to the total number of networks considered, minus the number of redundancy networks.
- the adequate model networks are then obtained by calculating the RMSE value for each network combination with a reduced network number.
- the network combination with the lowest RMSE value includes the model networks or predictors sought.
- Figure 3d shows the drift segmentation after Reduction step. Accordingly, the remaining predictors 2 and 5 fully describe the system.
- Another application of the invention is in the field of the analysis of physiological data which are characteristic of the course of sleep and wake modes of living beings.
- time series can e.g. B. segmented from EEG data.
- FIG. 4a shows the results of a conventional switching segmentation (top), a drift segmentation (middle) and a "manual" segmentation (bottom) of a medical specialist (sleep researcher) on the basis of empirical values using the example of an afternoon sleep of a healthy person.
- the switching and drift segmentations are carried out with eight networks (netl ... net ⁇ ) on single-channel EEG data x (t) (FIG. 4b).
- FIG. 4a like in FIG. 2, frames are drawn for reasons of clarity, which, in the case of the drift modes, illustrate the networks between which interpolation is carried out.
- the dotted line inside the frame shows the actual course.
- Manual segmentation is based on the observation of physiological signals (e.g.
- the modes Wl, W2 denote two wake modes with open or closed eyes and the modes S1, S2 each state of sleep. "n / A.” and “art.” relate to states or artifacts not considered.
- the switching segmentation shows a comparatively undifferentiated picture that is only roughly consistent with the other observations. For example,
- FIG. 4a shows that the method according to the invention can be used to automatically achieve detailed segmentations which were previously only accessible by observing complex feature images on the basis of broad experience and intuitions.
- This advantage can be used not only in medicine, but also in other areas in which large amounts of data are generated when describing complex dynamic systems. Such areas are physical, chemical and / or biological process engineering, geology, meteorology, climatology, language acquisition and the like. Like ..
- the system under consideration can be high-dimensional (10 or more dimensions).
- the invention allows the complexity of such a system to be reduced by considering lower dimensional modes and changing transitions between them.
- the use of predictive models for the Segmentation is invariant to changes in the amplitude of detected signals.
- the invention is used for the prediction or control of a system in such a way that, as described above, the actual state of the system, which possibly represents a mixture according to the result of the drift segmentation, is recorded from the past observation and knowledge of the current modes.
- the current state corresponds to a dynamic system f.
- the prediction means that the system f is applied to the current state x and from this the prediction for the immediately following state y results.
- the check means that the deviation from a TARGET state is determined from the current state and a suitable control strategy is derived from the deviation.
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Psychology (AREA)
- Heart & Thoracic Surgery (AREA)
- Psychiatry (AREA)
- Surgery (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Feedback Control In General (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
- Complex Calculations (AREA)
Abstract
Description
Claims
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP98948961A EP1015998A2 (de) | 1997-09-15 | 1998-09-11 | Verfahren zur erfassung zeitabhängiger moden dynamischer systeme |
US09/508,042 US7277831B1 (en) | 1997-09-15 | 1998-09-11 | Method for detecting time dependent modes of dynamic systems |
JP2000512152A JP2001516927A (ja) | 1997-09-15 | 1998-09-11 | 動的システムの時間依存モードの検出方法 |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DE19740565.7 | 1997-09-15 | ||
DE19740565A DE19740565A1 (de) | 1997-09-15 | 1997-09-15 | Verfahren zur Erfassung zeitabhängiger Moden dynamischer Systeme |
Publications (2)
Publication Number | Publication Date |
---|---|
WO1999014686A2 true WO1999014686A2 (de) | 1999-03-25 |
WO1999014686A3 WO1999014686A3 (de) | 1999-05-06 |
Family
ID=7842427
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/EP1998/005793 WO1999014686A2 (de) | 1997-09-15 | 1998-09-11 | Verfahren zur erfassung zeitabhängiger moden dynamischer systeme |
Country Status (5)
Country | Link |
---|---|
US (1) | US7277831B1 (de) |
EP (1) | EP1015998A2 (de) |
JP (1) | JP2001516927A (de) |
DE (1) | DE19740565A1 (de) |
WO (1) | WO1999014686A2 (de) |
Families Citing this family (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE10112038B4 (de) * | 2001-03-14 | 2008-06-12 | Testo Gmbh & Co Kg | Verfahren zur asynchronen,platzsparenden Datenerfassung innerhalb einer kontinuierlichen Messwertspeicherung |
DE10123572C1 (de) * | 2001-05-08 | 2003-01-23 | Senslab Gmbh | Verfahren und Vorrichtung zur automatischen Online-Analyse von Meßreihen sowie ein entsprechendes Computerprogramm-Erzeugnis und ein entsprechendes computerlesbares Speichermedium |
CN101802881B (zh) * | 2007-09-19 | 2012-08-15 | 皇家飞利浦电子股份有限公司 | 检测异常情况的设备和方法 |
EP2291790B1 (de) * | 2008-05-05 | 2020-04-29 | Exxonmobil Upstream Research Company | Modellierung von dynamischen geologischen systemen durch visualisierung und verschmälerung eines parameterraums |
US9775545B2 (en) | 2010-09-28 | 2017-10-03 | Masimo Corporation | Magnetic electrical connector for patient monitors |
WO2012050847A2 (en) | 2010-09-28 | 2012-04-19 | Masimo Corporation | Depth of consciousness monitor including oximeter |
WO2012076923A1 (en) * | 2010-12-08 | 2012-06-14 | Universite De Rouen | Method and system for automatic scoring of sleep stages |
WO2016057553A1 (en) | 2014-10-07 | 2016-04-14 | Masimo Corporation | Modular physiological sensors |
CN105629728B (zh) * | 2015-12-23 | 2018-07-17 | 辽宁石油化工大学 | 复杂动态网络的建模方法及模型控制器的设计方法 |
KR102324468B1 (ko) * | 2017-03-28 | 2021-11-10 | 삼성전자주식회사 | 얼굴 인증을 위한 장치 및 방법 |
US20190102718A1 (en) * | 2017-09-29 | 2019-04-04 | Oracle International Corporation | Techniques for automated signal and anomaly detection |
US11467803B2 (en) | 2019-09-13 | 2022-10-11 | Oracle International Corporation | Identifying regulator and driver signals in data systems |
Family Cites Families (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB8902645D0 (en) * | 1989-02-07 | 1989-03-30 | Smiths Industries Plc | Monitoring |
JP2964507B2 (ja) * | 1989-12-12 | 1999-10-18 | 松下電器産業株式会社 | Hmm装置 |
KR940009984B1 (ko) * | 1990-05-29 | 1994-10-19 | 미쓰비시덴키 가부시키가이샤 | 엘리베이터 제어장치 |
CA2040903C (en) * | 1991-04-22 | 2003-10-07 | John G. Sutherland | Neural networks |
JP2979711B2 (ja) * | 1991-04-24 | 1999-11-15 | 日本電気株式会社 | パターン認識方式および標準パターン学習方式 |
US5479571A (en) * | 1991-06-14 | 1995-12-26 | The Texas A&M University System | Neural node network and model, and method of teaching same |
EP0687369A1 (de) * | 1993-03-02 | 1995-12-20 | Pavilion Technologies Inc. | Methode und vorrichtung zur auswertung eines neuronalen netzes innerhalb von gewünschten betriebsparametergrenzen |
SE9304246L (sv) * | 1993-12-22 | 1995-06-23 | Asea Brown Boveri | Förfarande vid övervakning av multivariata processer |
US5659667A (en) * | 1995-01-17 | 1997-08-19 | The Regents Of The University Of California Office Of Technology Transfer | Adaptive model predictive process control using neural networks |
DE19530646C1 (de) * | 1995-08-21 | 1996-10-17 | Siemens Ag | Lernverfahren für ein rekurrentes neuronales Netz |
DE19530647C1 (de) * | 1995-08-21 | 1997-01-23 | Siemens Ag | Verfahren zur Aufbereitung einer Eingangsgröße für ein neuronales Netz |
DE19531967C2 (de) * | 1995-08-30 | 1997-09-11 | Siemens Ag | Verfahren zum Training eines neuronalen Netzes mit dem nicht deterministischen Verhalten eines technischen Systems |
DE19537010C2 (de) * | 1995-10-04 | 1997-10-02 | Siemens Ag | Lernverfahren und -anordnung zur Nachbildung eines dynamischen Prozesses |
US5748847A (en) * | 1995-12-21 | 1998-05-05 | Maryland Technology Corporation | Nonadaptively trained adaptive neural systems |
US5995868A (en) * | 1996-01-23 | 1999-11-30 | University Of Kansas | System for the prediction, rapid detection, warning, prevention, or control of changes in activity states in the brain of a subject |
US5796922A (en) * | 1996-03-29 | 1998-08-18 | Weber State University | Trainable, state-sampled, network controller |
EP0998720A4 (de) * | 1997-07-31 | 2005-03-23 | Univ California | Gerät und verfahren zur bild- und signalverarbeitung |
-
1997
- 1997-09-15 DE DE19740565A patent/DE19740565A1/de not_active Withdrawn
-
1998
- 1998-09-11 JP JP2000512152A patent/JP2001516927A/ja active Pending
- 1998-09-11 US US09/508,042 patent/US7277831B1/en not_active Expired - Fee Related
- 1998-09-11 EP EP98948961A patent/EP1015998A2/de not_active Withdrawn
- 1998-09-11 WO PCT/EP1998/005793 patent/WO1999014686A2/de active Application Filing
Non-Patent Citations (4)
Title |
---|
KOHLMORGEN J ET AL: "Analysis of Wake/Sleep EEG with Competing Experts" PROCEEDINGS OF THE 7TH INTERNATIONAL CONFERENCE ICANN '97 - ARTIFICIAL NEURAL NETWORKS, 8. - 10. Oktober 1997, Seiten 1077-1082, XP002093337 * |
POPIVANOV D ET AL: "DETECTION OF SUCCESSIVE CHANGES IN DYNAMICS OF EEG TIME SERIES: LINEAR AND NONLINEAR APPROACH" PROCEEDINGS OF THE 18TH ANNUAL INTERNATIONAL CONFERENCE OF THE IEEE ENGINEERING IN MEDICINE AND BIOLOGY SOCIETY, Bd. 4, 31. Oktober 1996 - 3. November 1996, Seiten 1590-1591, XP002093335 * |
PRDEY J ET AL: "A review for parametric modelling techniques for EEG analysis" MEDICAL ENGINEERING & PHYSICS, Bd. 18, Nr. 1, Januar 1996, Seiten 2-11, XP002093336 * |
See also references of EP1015998A2 * |
Also Published As
Publication number | Publication date |
---|---|
EP1015998A2 (de) | 2000-07-05 |
US7277831B1 (en) | 2007-10-02 |
JP2001516927A (ja) | 2001-10-02 |
WO1999014686A3 (de) | 1999-05-06 |
DE19740565A1 (de) | 1999-03-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP1589438B1 (de) | Verfahren und Vorrichtung zum Überwachen einer Konzentration eines Analyten im lebenden Körper eines Menschen oder Tieres | |
DE69526823T2 (de) | Gerät und Verfahren zum Analysieren von Information in Bezug auf physikalischen und geistigen Zustand | |
EP0856181B1 (de) | Verfahren und vorrichtung zur auswertung eines narkose- oder intensiv-eeg | |
DE69637337T2 (de) | Verbesserungen in Bezug auf physiologische Überwachung | |
DE69730811T2 (de) | Anlage zur Bilderkennung | |
DE69414752T2 (de) | Sprecherunabhängiges Erkennungssystem für isolierte Wörter unter Verwendung eines neuronalen Netzes | |
DE102017204628A1 (de) | Schlafzustandsvorhersagevorrichtung | |
EP0828225A1 (de) | Verfahren und Vorrichtung zum Auswerten von EEG-Daten | |
WO1999014686A2 (de) | Verfahren zur erfassung zeitabhängiger moden dynamischer systeme | |
Khan et al. | An expectation-maximization algorithm based Kalman smoother approach for event-related desynchronization (ERD) estimation from EEG | |
CN113662545B (zh) | 一种基于情绪脑电信号和多任务学习的人格评估方法 | |
EP1193688A2 (de) | Verfahren zur Ermittlung eines Eigenraums zur Darstellung einer Mehrzahl von Trainingssprechern | |
EP1145192B1 (de) | Anordnung miteinander verbundener rechenelemente, verfahren zur rechnergestützten ermittlung einer dynamik, die einem dynamischen prozess zugrunde liegt und verfahren zum rechnergestützten trainieren einer anordnung miteinander verbundener rechenelemente | |
Kohlmorgen et al. | Identification of nonstationary dynamics in physiological recordings | |
DE102020212515A1 (de) | Verfahren und Vorrichtung zum Trainieren eines maschinellen Lernsystems | |
DE4039648A1 (de) | Messwertverarbeitungssystem fuer ein biologisches objekt | |
EP1193689A2 (de) | Verfahren zur Ermittlung eines Eigenraums zur Darstellung einer Mehrzahl von Trainingssprechern | |
CN113180659A (zh) | 一种基于三维特征和空洞全卷积网络的脑电情感识别系统 | |
DE102020214050A1 (de) | Computerimplementiertes Verfahren und Vorrichtung zum Durchführen einer medizinischen Laborwertanalyse | |
EP0150352B1 (de) | Verfahren zur Bestimmung des Anfangs- und Endpunktes von geschlossenen, physiologischen Messsignalen | |
DE10047718A1 (de) | Verfahren zur Spracherkennung | |
Coyle et al. | Extracting features for a brain-computer interface by self-organising fuzzy neural network-based time series prediction | |
DE4209296A1 (de) | Wort- und wortkettenerkennungssystem auf der basis eines neuronalen netzes mit automatischer zellbildung und -komprimierung sowie nachlernverfahren | |
DE19610847C1 (de) | Verfahren zur Klassifikation einer meßbaren Zeitreihe, die eine vorgebbare Anzahl von Abtastwerten aufweist, beispielsweise eines elektrischen Signals, durch einen Rechner und Verwendung des Verfahrens | |
DE102020212514A1 (de) | Verfahren zur Ermittlung eines Ausgabesignals mittels eines maschinellen Lernsystems |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A2 Designated state(s): JP US |
|
AL | Designated countries for regional patents |
Kind code of ref document: A2 Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE |
|
DFPE | Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101) | ||
AK | Designated states |
Kind code of ref document: A3 Designated state(s): JP US |
|
AL | Designated countries for regional patents |
Kind code of ref document: A3 Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
WWE | Wipo information: entry into national phase |
Ref document number: 1998948961 Country of ref document: EP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 09508042 Country of ref document: US |
|
WWP | Wipo information: published in national office |
Ref document number: 1998948961 Country of ref document: EP |