Methods of analyzing multiple eds and computer readable media

Abstract

A method for operating a data processing system and a computer readable medium causing a data processing system to perform the method are disclosed. The method includes causing the data processing system to receive a plurality of first EDSs classified into a plurality of first clusters and a first RDS for each of the plurality of first clusters and displaying a first display for each of the first clusters and the RDS for each of the first clusters. The data processing system receives information from a user and performs a second clustering operation on selected clusters, the information specifying that one or more of the first clusters are to be further clustered to result in a specified number of second clusters, the specified one or more first clusters being classified into the second clusters. The method further includes displaying a second display, the second display including a plurality of second EDSs classified into the second clusters.

Description

[0001] The present application relates to a method for operating a data processing system to enable a user to analyze a plurality of EDSs and a computer readable medium containing instructions. BACKGROUND

[0002] Data recording systems are now capable of recording a substantial amount of data, such that the time to search the recorded data by serially reading the stored data becomes substantial. Often data sets of over 1 terabyte are recorded. These data sets can be the result of monitoring a stream of signal values, images, or other quantities of interest. In many applications of interest, the data stream is composed of multiple signals of interest separated from each other by background data. In U.S. Patent Application No. 16 / 373,343, filed April 2, 2019, a system for pre-processing such a data stream without detailed prior knowledge of the signals of interest is disclosed. The pre-processing produces a database comprising clusters of similar signals. Each cluster is characterized by a representative element of the cluster, the number of elements in the cluster, and other useful information. While this system significantly reduces the amount of data that must be understood in order to understand the content of the data stream, there remains a challenge to understand the data stream in terms of underlying clusters that are important to the user. SUMMARY

[0003] The present application includes a method for operating a data processing system and a computer readable medium that causes a data processing system to perform the method. The method operates on a data processing system having a user interface and a display. The method includes causing the data processing system to receive a plurality of first extracted data segments (EDSs) classified into a plurality of first clusters and one first reference data segment (RDS) for each of the plurality of first clusters, and displaying on a region of the display a first display for each of the plurality of first clusters and the RDS for each of the plurality of first clusters. The method further includes receiving information from a user, designating one or more of the first clusters to be further clustered to result in a designated number of second clusters, and performing a second clustering operation of the one or more first clusters that are to be classified into the second clusters. The method further includes displaying on the first display region a second display, the second display comprising a plurality of second EDSs classified into the second clusters as a result of the second clustering operation.

[0004] In one aspect, the first display includes a number of EDSs belonging to each of the plurality of first clusters.

[0005] In another aspect, the second clustering operation includes a hierarchical clustering method.

[0006] In another aspect, the first display includes, for each of the plurality of first clusters, a first RDS of each of the plurality of first clusters and a number of EDS belonging to each of the plurality of first clusters, including arranging and displaying the first RDS and the number of EDS in a format of a plurality of tiles horizontally for the plurality of first clusters.

[0007] In another aspect, each of the tiles is characterized by a horizontal display range and a vertical display range, and wherein the horizontal display range and the vertical display range are set individually for each of the tiles such that the first RDS displayed in the tile substantially occupies all of the horizontal display range and the vertical display range.

[0008] In another aspect, the first display includes, for each of the plurality of first clusters, a first RDS of each of the plurality of first clusters and a number of EDS belonging to each of the plurality of first clusters, including arranging and displaying the first RDS and the number of EDS in a format of a list of tiles vertically for the plurality of first clusters.

[0009] In another aspect, each of the list of tiles has a same horizontal proportion.

[0010] In another aspect, each of the list of tiles has a vertical proportion optimized for the RDS associated with the tile.

[0011] In another aspect, the receiving information about the first cluster selected from among the plurality of first clusters as a next classification target includes displaying RDSs of all the selected first clusters on a second display area so as to overlap with each other with a common amplitude scale and a common time scale.

[0012] In another aspect, the receiving information about the first cluster selected from among the plurality of first clusters as a next classification target includes displaying any one of EDS belonging to the selected first cluster on a third display area.

[0013] In another aspect, displaying any one of the EDS belonging to all the selected first clusters on a third display area includes displaying the EDS belonging to all the selected first clusters one by one by using a first control button to input a time order of the EDS.

[0014] In another aspect, the step of displaying each of the EDSs belonging to all of the selected first clusters in a time sequence further comprises repeatedly displaying each of the EDSs belonging to all of the selected first clusters in a continuous display when a second control button is pressed.

[0015] In another aspect, each of the EDSs comprises a vector-valued function of time.

[0016] In another aspect, each of the EDSs comprises a plurality of images as a function of time. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 A data structure used in one embodiment of the application is illustrated.

[0018] Figure 2 Components of a data analysis system according to one embodiment of the application are illustrated.

[0019] Figure 3 A basic screen for communicating with a user in the EDS display unit 208 is illustrated.

[0020] Figure 4 An embodiment of the cluster selection sub-display 402 in which a list mode is used is illustrated.

[0021] Figure 5 An embodiment of the application in which the EDS sub-display 412 is replaced by a sub-display 712 is illustrated.

[0022] Figure 6 Possible screen displays for the EDS are illustrated. DETAILED DESCRIPTION

[0023] Consider a set of objects, where each object is characterized by a signature. For example, consider the objects to be sequences of recorded signal values that satisfy some extraction condition that is used to identify objects in a recorded stream. The signature corresponding to an object can be the signal values themselves, some transform of the signal values such as the coefficients of a Fourier transform of the signal as a function of time, or some other transform of the signal values. In general, the signature corresponding to an object will be a multivalued quantity; thus, the signature can be viewed as a vector with multiple components.

[0024] In many cases, the goal of the clustering operation process is to find clusters of objects in a coordinate system in which each axis corresponds to a different component of the signature components. Each object can be viewed as a point in this coordinate system. Typically, a relationship is defined between two signatures that provides a measure of the similarity of the two signatures. A cluster is typically defined as a group of objects that are similar to each other as judged by the similarity measure.

[0025] For example, the similarity measure can be the distance between two signatures in the signature space. All objects having a similarity measure less than some predetermined threshold from one of the objects are defined to be in the same cluster.

[0026] The manner in which the present invention provides its advantages can be more easily understood with reference to a data recording system in which signals in an input data channel are digitized and stored on a memory device such as a disk drive. The data stream can be viewed as containing signals of interest defined by an "extraction algorithm" that identifies a sequence of signal values of interest as an EDS.

[0027] Typically, a user of the recorded data needs to be able to understand the various signals in the data and retrieve signals of interest. For the purposes of this discussion, it will be assumed that the user does not have detailed knowledge of all of the signals in the data stream of interest. In addition, it will be assumed that the number of data stream signals is too large for the user to view one data stream signal at a time. Thus, the user needs to be able to understand the important characteristics of the signals without viewing the entire data stream. To this end, it is effective to define clusters of similar signals. By examining representative members of such clusters, the user can better understand the recorded signals and specify the parameters needed to retrieve signals of interest, as well as select clusters based on viewing the combination of signals corresponding to the cluster.

[0028] The present invention provides the user with a tool that allows the user to define clusters in a set of signals that have been recorded based on a similarity algorithm. The similarity algorithm computes a similarity measure related to the similarity between two signals. Algorithms for clustering objects based on similarity measures are known in the art. Unfortunately, the computational effort inherent in applying many of these algorithms is on the order of N2or higher, where N is the number of signals. Given that a few terabyte data stream can have more than a few million signals, it is generally not practical to cluster the signals in a few minutes while the user is exploring the recorded signals unless some mechanism is provided for reducing the number of signals that must be clustered at one time. 2

[0029] ​As will be explained in greater detail below, small clusters of signals of interest are detected during the recording process or during playback of previously recorded data streams. These small clusters are then combined using the present application to provide larger clusters that are consistent with the clusters of signals in the input data stream. The small clusters are constructed without requiring a detailed predetermined description of the signals to be clustered. Ideally, each of these clusters contains a portion of a single cluster of underlying signals present in the input stream. As discussed below, each cluster begins with an observed signal in the input stream. The size of the cluster is determined by a similarity algorithm that includes a threshold that determines whether a second signal is to be included in the same cluster as a first signal. The manner in which the clusters are combined or the manner in which the clusters are broken down into smaller clusters will be discussed in greater detail below.

[0030] For simplicity of the following discussion, it will be assumed that the signals of interest are raw data values and that the "signature" for each signal of interest is a vector of signal values. Other cases will be discussed in greater detail below. It will be assumed that the data stream consists of individual signals separated by regions that do not include segments of data of interest. Segments of the data stream that satisfy the extraction algorithm will be referred to as EDSs.

[0031] Ideally, each EDS will contain data samples corresponding to one signal of interest without any background samples. However, it is often necessary to identify EDSs in a short period of time, and thus this requires a limitation on the extraction algorithm. There are many computationally efficient methods for detecting the beginning of some signal that is different from the background as known in the art. For example, the extraction algorithm can look for a rising edge or a falling edge. However, detecting the point at which a signal returns to the background level is more computationally complex, especially in the presence of noise. Thus, the extraction algorithm is prioritized as the algorithm for defining EDSs in which the end of a signal of interest is defined as a fixed number of samples from the beginning of the signal. If two signals are in fact identical, the EDSs for the two signals will still match. Thus, in one exemplary embodiment, it is assumed that the data stream consists of individual signals separated by regions that do not include segments of data of interest. If this approximation interferes with the final clustering operation, the EDSs can be retrieved from long term storage and the clustering operation can be performed based on the more accurate end of the signal.

[0032] The EDSs are also defined by a "similarity metric" by the similarity algorithm. The similarity metric reflects the degree of similarity between any two EDSs. The similarity metric allows the system to group the extracted data segments into clusters of EDSs that are similar to each other. In one aspect, the similarity algorithm includes a threshold. If the similarity metric has a predetermined relationship to the threshold, then two EDSs are defined as similar to each other. For example, if the similarity metric is less than the threshold, then the two EDSs can be defined as similar to each other. EDSs that are similar to each other are grouped in a cluster.

[0033] When a new EDS is found, the system determines whether the EDS is part of a cluster that has already been found. If the EDS is part of an existing cluster, the existing cluster is updated to reflect the addition of the new EDS. If the EDS is not sufficiently similar to any existing cluster, a new cluster is defined and the EDS is added to the cluster.

[0034] Each cluster is represented by an RDS. If a new EDS is similar to an existing RDS, the new EDS is marked as belonging to the cluster represented by the RDS. If the new EDS is not similar to any existing RDS, a new cluster is defined for the EDS and the EDS becomes the RDS for the new cluster.

[0035] Reference is now made to Figure 1 which illustrates a data structure used in one embodiment of the present application. The data structure 100 includes a block for each of the clusters that have been extracted as described above. Exemplary blocks are shown at 101-103. Each block includes the RDS associated with the corresponding cluster and other information about the cluster such as the number of EDS in the cluster and information about the similarity metric used to generate the cluster. Each block also includes a plurality of EDS that were found to be similar to the RDS. Exemplary EDS are shown at 104-106. The EDS can be stored as the corresponding string of data values or as a pointer to a storage location containing the original data values.

[0036] Although the above example provides a method for obtaining the clusters of EDS and the RDS associated with each cluster, other methods for providing a representative EDS corresponding to each cluster can be utilized. The RDS described above is the first EDS that cannot be assigned to any existing cluster. However, once a cluster has been defined, the RDS selection for that cluster can be determined in other ways. For example, the RDS can be selected by statistical methods involving, for example, calculating the similarity between all EDS belonging to each cluster by using, for example, an evaluation function such as the Euclidean distance, and selecting an RDS as the representative EDS by taking the median of these results. For the purposes of the present disclosure, any method for defining the RDS for a cluster of EDS can be utilized.

[0037] Reference is now made to Figure 2which illustrates the components of a data analysis system according to one embodiment of the present application. Analysis system 200 includes an EDS acquisition unit 202, an EDS analysis unit 204, an EDS classification unit 206, and an EDS display unit 208. EDS acquisition unit 202 is configured to acquire an input data stream from which EDS is to be extracted. The source of the data stream can be a measurement device or a data logger that is playing back a previously recorded data stream. EDS acquisition unit 202 extracts EDS from the acquired input, classifies the EDS into preliminary clusters by means of EDS classification unit 206, and stores the results in a memory structure shown in Figure 1 In one aspect of the present application, extraction conditions are used to identify data segments to be EDS.

[0038] EDS display unit 208 operates to receive data from EDS analysis unit 204 and display the data on a display 210 that is part of EDS display unit 208. EDS display unit 208 is also configured to receive user input that specifies the manner in which data is to be displayed and instructions regarding various processing steps, including re-clustering operations on EDS in a current cluster.

[0039] For example, when a classification is needed for further narrowing of a classification in response to a classification result, the user informs EDS display unit 208 of new classification parameters. EDS classification unit 206 then receives the instructions, executes the instructions, and updates information about clusters within a memory area to narrow the results.

[0040] In one exemplary embodiment, EDS are initially classified by EDS acquisition unit 202. Subsequently, one or more of the clusters can be processed in response to user input to find a better RDS for the cluster, to split the cluster into multiple clusters, or to combine clusters. In addition, clusters can be hierarchically combined using the same similarity algorithm of different algorithms. Also, EDS can be processed to provide a signature for each EDS in a cluster and reclassified using the signature and different similarity algorithms or thresholds.

[0041] Reference is now made to Figure 3 which illustrates a basic screen used in EDS display unit 208 for communication with a user. Screen 400 has three sub-displays shown at 402, 410, and 412. Cluster selection sub-display 402 will be referred to as the cluster selection sub-display. Cluster sub-display 410 will be referred to as the selected cluster sub-display, and EDS sub-display 412 will be referred to as the EDS display.

[0042] The cluster selection sub-display 402 includes multiple panes, one for each cluster in the current cluster set. The current cluster set can include new clusters generated from the preliminary clusters. An exemplary pane is labeled at 408. In this example, each pane includes a display of the RDS associated with that cluster, a checkbox that the user can use to select that cluster, and information about the number of EDS in the cluster.

[0043] The particular display format for the cluster selection sub-display 402 is specified in the drop-down menu 404. In this example, a tiled display is selected. In a tiled display, the pane for each cluster includes a display of the RDS for that cluster, with the display providing the maximum detail for the RDS within the space allocated for each pane in a format that is optimized. Thus, the vertical and horizontal axes of the graph for the RDS are set to be substantially equal to the horizontal and vertical ranges of the RDS. As a result, it is not necessary to compare two RDS by comparing the displays in their respective panes. Other display formats that allow such direct comparison will be discussed below.

[0044] In general, a tiled display has multiple rows of panes and multiple columns of panes. Each row has multiple panes. If the number of clusters requires more rows than are available in the display space, the rows of panes can be scrolled. The cluster selection sub-display also includes a menu 406 that allows the user to specify the count of the number of clusters to be displayed in the sub-display without having to scroll the rows of the sub-display.

[0045] As noted above, each pane in the cluster selection sub-display 402 includes a checkbox that allows the user to select the cluster corresponding to that pane for further processing. Each of the selected clusters is displayed in the selected cluster sub-display 410. The RDS corresponding to each of the selected clusters is displayed in an overlaid display in which the common overlaid display horizontal axis is the same for all of the RDS, and the vertical scale is chosen to be sufficient to allow all of the selected RDS to be displayed in the sub-display. Thus, the user can better compare the selected clusters.

[0046] In one aspect of the application, the user can use the hierarchical clustering to specify that the selected clusters be combined into a specified number of clusters. In one aspect, a new RDS is selected for each of the newly created clusters. The newly created clusters are then added to the list of clusters displayed in the cluster selection sub-display 402, and the clusters that were combined to provide the new clusters are removed.

[0047] In another aspect of the application, a single cluster can be selected, and the EDS classification unit 206 can receive instructions from the user to reclassify the EDS of that cluster into multiple clusters using a new similarity metric. For example, the initial similarity algorithm can be used with a more stringent threshold.

[0048] In another reclassification example, the EDSs of selected clusters can be analyzed to find more accurate endpoints than those used in the initial data segment extraction. As noted above, the extraction algorithm selects a predetermined number of data samples relative to a trigger that begins the extraction. In this aspect, each EDS is examined to determine if a more accurate estimate of the number of samples that are actually in the segment of interest can be determined. The more accurate estimate is then used to define the endpoints of the EDS. This analysis can result in clusters in which the EDSs have different lengths. The clusters can then be reclassified into a plurality of clusters, each new cluster having a range of lengths, with different clusters having different ranges.

[0049] The display described above operates on the RDS associated with each of the clusters. Although the RDS is also an EDS, other EDSs in the cluster can be important for the user to understand the data stream. The EDS sub-display 412 allows the user to view the individual EDSs belonging to the selected cluster. In this embodiment, the EDSs are displayed one at a time. Each EDS has a unique identifier. In this example, the identifier is a timestamp representing the time at which the EDS occurs in the data stream. Thus, the EDSs are in order. By specifying an EDS using its identifier, the EDS can be displayed in the EDS sub-display.

[0050] The detailed EDS is displayed on pane 414 of the EDS sub-display 412. For the currently displayed EDS, the EDS position indicator 418 displays information about the start point and end point relative to the time position in the original data stream (0 days 00:00:15.114, 474,000), as well as the number of data points in the EDS (0 days 00:00:15.114, 474,000), and the associated cluster name (C5). The time bar 424 and the current position indicator 426 indicate the position of the currently displayed EDS in the time bar 424 by the current position indicator 426. The left triangle arrow icon 420 and the right triangle arrow icon 422 are move instruction buttons that will be used to move the position of the current position indicator 426 by one EDS. Further, on the top left box 416 of the sub-display, the operation control buttons are provided from the left in a fixed order, namely "reverse playback", "back one EDS", "stop", "forward one EDS", and "playback", and these buttons can be used to display the EDS at the desired position. On the detailed EDS sub-display 412, any one of the EDSs belonging to any selected cluster is displayed, and the operation control buttons represented by the box 416 can be used to display "forward one EDS", "back one EDS", "repeat playback of all EDSs", or "repeat reverse playback of all EDSs".

[0051] Icon 428 specifies whether a single or dual screen display mode is active for EDS sub-display 412. The dual screen display mode is discussed in detail below.

[0052] In the embodiment shown in Figure 3 , region 408 is populated by a tiled display of clustered RDSs. In a tiled display, the horizontal and vertical axes of the plot are chosen to maximize the detail of the RDSs. Thus, the horizontal and vertical scales of the different plots are different. While this arrangement provides maximum visibility of each RDS in the available space, in some applications the differences in scale can make it difficult for the user to fully understand the different RDSs. To address this problem, the present invention provides a "list mode" in which the horizontal scale in the different plots is the same.

[0053] Reference is now made to Figure 4 , which illustrates an embodiment of clustered selection sub-display 402 in which the list mode is used. List display 600 replaces pane 402 shown in Figure 3 . Classification results display screen 602 utilizes a list display selected by choosing the "list" mode from drop-down menu 604. In a format similar to that described above with reference to Figure 3 , representative plots of each cluster and the number of EDSs of the elements of the respective cluster are displayed vertically in rows. Since all nine clusters cannot be displayed simultaneously on the list display, a scroll bar 607 is displayed on the right side of display 608. The display of each cluster includes a check box for checking whether the cluster is selected, an RDS, a length indicating the length of the RDS, and the number of EDSs (Refs) contained in the cluster. When the list mode is selected, Figure 3 , the remaining sub-displays operate in a similar manner, and thus, these sub-displays are not shown in Figure 4 .

[0054] In the above described embodiments, detailed EDS sub-display 412 displays a single view of a single EDS. However, embodiments in which the EDS sub-display has multiple displays can also be constructed. Reference is now made to Figure 5 , which illustrates one embodiment of the present invention in which EDS sub-display 412 is replaced by sub-display 712. For simplicity of the drawing, only the EDS sub-display is shown, the remaining sub-displays operate as described above.

[0055] This dual pane mode is selected by clicking on icon shown at 716. In this dual pane display mode, pane 714 displays the plots of the selected EDS in a format similar to that described above with reference to Figure 3The currently selected EDS is described in a manner selected. Pane 720 displays a detailed view of the displayed portion 716 in pane 714, which is selected by the user using a pointing device associated with the user interface.

[0056] In another aspect of the application, the user can repeatedly view selected EDSs in an overlay display, which allows the user to view the EDSs in chronological order in a manner that allows the user to watch each EDS "evolve" into the next EDS in the sequence. In this embodiment, the user selects a cluster containing the EDSs to be displayed. The individual EDSs of the selected cluster or clusters are then displayed. This display mode is similar to a persistence display in an oscilloscope.

[0057] For purposes of this disclosure, a "persistence display" is defined as a display with the following characteristics. All selected EDSs are displayed using the same horizontal and vertical axes, as if a single, overlaid display were created. However, each EDS is "drawn" on this display screen for a finite period of time, during which the EDS is initially displayed in a high intensity mode for a first period of time, and then gradually fades away via a lower intensity display until it disappears from the screen. The time required to sequentially begin each of the individual EDS displays is referred to as the display period. The display period is divided into N display intervals, where N is the number of EDSs to be displayed. At the beginning of each interval, the next EDS in the sequence begins its display. Any given EDS is visible on the screen for longer than a display interval; thus, the display evolves from one EDS to the next. In one aspect, the EDSs are presented in the order of their occurrence in the original data stream, and each EDS is visible for at least three display intervals.

[0058] In other embodiments, the user can select the order in which the EDSs are displayed. For example, if the various clusters are further clustered using a hierarchical clustering, the RDSs for each cluster can be displayed in an order that represents the evolution of the RDSs along the hierarchical clustering tree.

[0059] In the above described embodiments, the input data stream is scalar in nature. That is, it consists of a single value on each clock cycle. However, the teachings of the present invention can be applied to vector input data streams. In a vector value input data stream, there are multiple scalar values for each point in time. Such a data stream can be presented as multiple input data channels, where each channel is processed by an ADC to provide an input vector on each clock cycle. In another example, a multiple data point for each point in time is generated by transforming a scalar value input stream. For example, a time domain input signal can be converted to a frequency domain signal by filtering the time domain input stream using a band pass filter to generate the amplitude of the frequency component as a function of time. Multiple such filters will provide a three dimensional input stream, where the first dimension is time, the second dimension is frequency, and the third dimension is the amplitude of the frequency component. The opening trigger circuit defining the new EDS can operate on one or more of the channels.

[0060] The user interface of the present invention can be applied to such vector value input streams, providing a suitable format that can be defined to display the resulting EDS and RDS. In the case of a vector value data stream having two components for each point in time, a three dimensional display can be utilized, where the first axis is time, the second axis is one of the two components, and the third axis is the third component. The third axis can be displayed as a density value, such that the resulting display resembles a photograph with gray levels representing the third value. Alternatively, the display can be a perspective view of a three dimensional surface. In the case of a three component vector, a fourth component can be encoded as a color, such that each point in the picture has a color and intensity.

[0061] In another example, a data stream consisting of three dimensional objects, such as regular images, the time axis can be presented as a sequence of images. Referring to Figure 6 which illustrates a possible screen display for such an EDS. The EDS 800 will then be a series of consecutive images 801 to 804 that satisfy some extraction condition, such as a scene change. The zoomed in display of such a string of images can be a single image or a portion of a single image, as well as a subset of the string of images. It should be noted that the four dimensional display discussed above can also be implemented as a string of three dimensional displays.

[0062] The similarity function between two images used to cluster the images can be implemented as a cross correlation between the images or a cross correlation between the two images, where one image is shifted with respect to the other. It should also be noted that a sequence of images is a special case of a vector value input data stream, where each "vector" has NxM components, which are the pixels of an NxM pixel image.

[0063] All other features described above for the embodiment where the RDS is displayed as a two dimensional graph can be utilized by this three dimensional display instead of the two dimensional display.

[0064] Referring again to Figure 2 Although the EDS classification unit 206 is shown as being separate from the EDS analysis unit 204, the EDS classification unit 206 can be implemented as part of a general purpose computer, as can the EDS display unit 208. The memory block storing the clustering information can be implemented in the memory of the general purpose computer or in an external storage device connected to the computer.

[0065] If the data rate is sufficiently low, the EDS acquisition unit 202 can be implemented as an analog input channel to a general purpose computer. In the case of a pre-recorded data stream in digital form, the EDS acquisition unit 202 is preferably an input channel of a general purpose computer. Alternatively, the EDS acquisition unit 202 can be implemented in separate hardware similar to the input section of an oscilloscope. This type of input hardware can include a number of sampling and digitizing circuits operating in parallel and thus be capable of a very high data input rate.

[0066] A general purpose computer can also advantageously be implemented as a multi-processor. In various clustering operations and re-clustering operations, the EDSs can be matched to each other in a process that can be speeded up by utilizing a multi-processor, since the matching result between two EDSs can be carried out in parallel with the matching between two other EDSs without interfering with the matching of the two initial EDSs. The multi-processor can be a regular multi-core computer or a graphics processing board with thousands of cores.

[0067] The present invention also includes a computer readable medium having stored therein instructions for causing a data processing system to perform the method of the present invention. A computer readable medium is defined as any medium or means that comprises a program of instructions that is readable by a computer or data processing system and that causes the computer or data processing system to perform a method as defined by the instructions. Such a medium or means can be a non-transitory medium, such as a computer memory device, that stores information in a format readable by a computer or data processing system.

[0068] The present application relates to the following technical solution:

[0069] 1. A method for operating a data processing system to enable a user to analyze a plurality of EDSs, the data processing system having a user interface and a display, the method comprising:

[0070] causing the data processing system to receive a plurality of first EDSs classified into a plurality of first clusters and a first RDS for each of the plurality of first clusters;

[0071] displaying a first display on a first display area of the display, the first display for each of the plurality of first clusters and a RDS for each of the plurality of first clusters; receiving information from a user, the information specifying that one or more of the plurality of first clusters are to be further clustered to result in a specified number of second clusters, the specified one or more first clusters to be classified as the second clusters;

[0072] performing a second clustering operation of the one or more first clusters; and

[0073] displaying a second display on the first display area, the second display including a plurality of second EDS,

[0074] the plurality of second EDS classified as the second clusters as a result of the second clustering operation.

[0075] 2. The method of clause 1, wherein the first display includes a number of EDS belonging to each of the plurality of first clusters.

[0076] 3. The method of clause 1, wherein the second clustering operation includes a hierarchical clustering method.

[0077] 4. The method of clause 1, wherein, for each of the plurality of first clusters, the first display includes a first RDS for each of the plurality of first clusters and a number of EDS belonging to each of the plurality of first clusters, including arranging and displaying the first RDS and the number of EDS in a format of a plurality of tiles horizontally for the plurality of first clusters.

[0078] 5. The method of clause 4, wherein each of the tiles is characterized by a horizontal display range and a vertical display range, and wherein the horizontal display range and the vertical display range are individually set for each of the plurality of tiles such that the first RDS displayed in the tile substantially occupies all of the horizontal display range and the vertical display range.

[0079] 6. The method of clause 1, wherein, for each of the plurality of first clusters, the first display includes a first RDS for each of the plurality of first clusters and a number of EDS belonging to each of the plurality of first clusters, including arranging and displaying the first RDS and the number of EDS in a format of a plurality of list-formatted tiles vertically for the plurality of first clusters.

[0080] 7. The method of clause 6, wherein each of the plurality of list-formatted tiles has a same horizontal proportion.

[0081] 8. The method of clause 7, wherein each of the plurality of list-formatting tiles has a vertical proportion optimized for the RDS associated with that tile.

[0082] 9. The method of clause 1, wherein receiving information about the plurality of first clusters selected from among the plurality of first clusters as next classification targets includes displaying RDSs of all selected first clusters on a second display area so as to overlap with each other with RDSs having a common amplitude proportion and with RDSs having a common time proportion.

[0083] 10. The method of clause 1, wherein receiving information about the first clusters selected from among the plurality of first clusters as next classification targets includes displaying any of the plurality of EDSs belonging to the selected first clusters on a third display area.

[0084] 11. The method of clause 10, wherein the displaying any of the EDSs belonging to all of the selected first clusters on the third display area includes displaying the EDSs belonging to all of the selected first clusters one by one by using a first control button that displays the EDSs in an order determined by a time at which the EDSs are input.

[0085] 12. The method of clause 10, wherein the displaying the EDSs belonging to all of the selected first clusters one by one by using the first control button further includes repeatedly displaying the EDSs belonging to all of the selected first clusters in a continuous display manner when a second control button is pressed.

[0086] 13. The method of clause 1, wherein each of the EDSs includes a vector-valued function of time.

[0087] 14. The method of clause 13, wherein each of the EDSs includes a plurality of images varying over time.

[0088] 15. A computer-readable medium containing instructions, which when loaded into a data processing system, cause the data processing system to perform a method that enables a user to analyze a plurality of EDSs, the method comprising:

[0089] causing the data processing system to receive a plurality of first EDSs classified into a plurality of first clusters and a first RDS for each of the plurality of first clusters;

[0090] displaying a first display on a first display area of a display associated with the data processing system, the first display for each of the plurality of first clusters and RDS for each of the plurality of first clusters;

[0091] receiving information from a user, the information specifying that one or more of the plurality of first clusters are to be further clustered to result in a specified number of second clusters, the specified one or more first clusters to be classified as the second clusters;

[0092] performing a second clustering operation of the one or more first clusters; and

[0093] displaying a second display on the first display area, the second display including a plurality of second EDS, the plurality of EDS respectively classified as the second clusters as a result of the second clustering operation.

[0094] 16. The computer readable medium of clause 15, wherein, for each of the plurality of first clusters, the first display includes a first RDS of each of the plurality of first clusters and a number of EDS belonging to each of the plurality of first clusters, including arranging and displaying the first RDS and the number of EDS in a plurality of tiles horizontally for the plurality of first clusters.

[0095] 17. The computer readable medium of clause 15, wherein the receiving information regarding a first cluster selected from the plurality of first clusters as a next classification target includes displaying RDS of all selected first clusters on a second display area so as to overlap with each other with RDS having a common amplitude scale and a common time scale.

[0096] 18. The computer readable medium of clause 15, wherein the receiving information regarding a first cluster selected from the plurality of first clusters as a next classification target includes displaying any of the EDS belonging to the selected first cluster on a third display area.

[0097] 19. The computer readable medium of clause 15, wherein each of the EDS includes a vector valued function of time.

[0098] 20. The computer readable medium of clause 19, wherein each of the EDS includes a plurality of images varying over time.

[0099] The above-described embodiments of the application have been provided to demonstrate various aspects of the application. It should be understood, however, that the various aspects of the application illustrated in the different embodiments can be combined in other embodiments to provide other embodiments of the application. In addition, various modifications to the specifically described embodiments can be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other embodiments without departing from the scope of the application. Thus, the present application is not intended to be limited to the embodiments described herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for operating a data processing system to enable a user to analyze a plurality of EDSs, the data processing system having a user interface and a display, the method comprising: causing the data processing system to receive a plurality of first EDSs classified into a plurality of first clusters and a first RDS for each of the plurality of first clusters, the first RDS for each of the plurality of first clusters being one of the EDSs in one of the plurality of first clusters, each of the plurality of first EDSs being a segment of data automatically extracted from a data stream, the data stream comprising an ordered sequence of data values, the ordered sequence of data values being based on the data values in the data stream; displaying a first display on a first display area of the display, the first display for each of the plurality of first clusters and the RDS for one of the plurality of first clusters; receiving information from the user, the information specifying one or more of the plurality of first clusters to be further clustered to result in a specified number of second clusters, the specified one or more first clusters to be classified into the second clusters; wherein receiving information regarding the plurality of first clusters selected from among the plurality of first clusters as the next classification target includes displaying the RDSs of all selected first clusters on a second display area so as to overlap with each other with RDSs having a common amplitude scale and with RDSs having a common time scale; performing a second clustering operation of the one or more of the plurality of first clusters to obtain the specified number of second clusters; and displaying a second display for each of the second clusters on the first display area, each second display including an RDS representing the second cluster, the RDS being one of the EDSs in the second cluster.

2. The method of claim 1, wherein, the first display includes a number of EDSs belonging to each of the plurality of first clusters.

3. The method of claim 1, wherein, the second clustering operation includes a hierarchical clustering method.

4. The method of claim 1, wherein, for each of the plurality of first clusters, the first display includes a first RDS of each of the plurality of first clusters and a number of EDSs belonging to each of the plurality of first clusters, including arranging and displaying the first RDS and the number of EDSs in a format of a plurality of tiles horizontally for the plurality of first clusters.

5. The method of claim 4, wherein, each of the plurality of tiles is characterized by a horizontal display range and a vertical display range, and wherein the horizontal display range and the vertical display range are individually set for each of the plurality of tiles such that the first RDS displayed in the tile substantially occupies all of the horizontal display range and the vertical display range.

6. The method of claim 1, wherein, for each of the plurality of first clusters, the first display includes a first RDS of each of the plurality of first clusters and a number of EDSs belonging to each of the plurality of first clusters, including arranging and displaying the first RDS and the number of EDSs in a format of a plurality of list-formatted tiles vertically for the plurality of first clusters.

7. The method of claim 6, wherein, Each of the plurality of list-formatting tiles has the same horizontal scale.

8. The method of claim 7, wherein, Each of the plurality of list-formatting tiles has a vertical scale optimized for the RDS associated with that tile.

9. The method of claim 1, wherein, Receiving information regarding the first cluster selected from among the plurality of first clusters as a next classification target includes displaying any of the plurality of EDS belonging to the selected first cluster on a third display area.

10. The method of claim 9, wherein, The displaying on the third display area any of the EDS belonging to all of the selected first clusters includes displaying the EDS belonging to all of the selected first clusters one by one by using a first control button that displays the EDS in an order determined by the time at which the EDS were input.

11. The method of claim 10, wherein, The displaying the EDS belonging to all of the selected first clusters one by one by using a first control button further includes repeatedly displaying the EDS belonging to all of the selected first clusters in a continuous display manner when a second control button is pressed.

12. The method of claim 1, wherein, Each of the EDS includes a vector-valued function of time.

13. The method of claim 12, wherein, Each of the EDS includes a plurality of images that vary over time.

14. A computer-readable medium containing instructions which, when loaded into a data processing system, cause the data processing system to perform a method that enables a user to analyze a plurality of EDS, the method comprising: causing the data processing system to receive a plurality of first EDS classified into a plurality of first clusters and a first RDS for each of the plurality of first clusters, the first RDS for each of the plurality of first clusters being an EDS in one of the plurality of first clusters, each of the plurality of first EDS being a data segment automatically extracted from a data stream, the data stream comprising an ordered sequence of data values, the ordered sequence of data values being based on the data values in the data stream; displaying a first display on a first display area of a display, the first display being for each of the plurality of first clusters and the RDS for one of the plurality of first clusters; receiving information from a user, the information specifying one or more of the plurality of first clusters to be further clustered to result in a specified number of second clusters, the specified one or more first clusters to be classified into the second clusters; wherein receiving information regarding the plurality of first clusters selected from among the plurality of first clusters as a next classification target includes displaying the RDS of all of the selected first clusters on a second display area so as to overlap each other with RDSs having a common amplitude scale and having a common time scale; performing a second clustering operation of the one or more of the plurality of first clusters to obtain the specified number of second clusters; and displaying a second display for each of the second clusters on the first display area, each second display including an RDS representing the second cluster, the RDS being one of the EDS in the second cluster. Each of the EDS includes a vector-valued function of time. Each of the EDS includes a plurality of images that vary over time.

14. A computer-readable medium containing instructions which, when loaded into a data processing system, cause the data processing system to perform a method that enables a user to analyze a plurality of EDS, the method comprising: causing the data processing system to receive a plurality of first EDS classified into a plurality of first clusters and a first RDS for each of the plurality of first clusters, the first RDS for each of the plurality of first clusters being an EDS in one of the plurality of first clusters, each of the plurality of first EDS being a data segment automatically extracted from a data stream, the data stream comprising an ordered sequence of data values, the ordered sequence of data values being based on the data values in the data stream; displaying a first display on a first display area of a display, the first display being for each of the plurality of first clusters and the RDS for one of the plurality of first clusters; receiving information from a user, the information specifying one or more of the plurality of first clusters to be further clustered to result in a specified number of second clusters, the specified one or more first clusters to be classified into the second clusters; wherein receiving information regarding the plurality of first clusters selected from among the plurality of first clusters as a next classification target includes displaying the RDS of all of the selected first clusters on a second display area so as to overlap each other with RDSs having a common amplitude scale and having a common time scale; performing a second clustering operation of the one or more of the plurality of first clusters to obtain the specified number of second clusters; and displaying a second display for each of the second clusters on the first display area, each second display including an RDS representing the second cluster, the RDS being one of the EDS in the second cluster.

15. The computer readable medium of claim 14, wherein, The first display includes, for each of the plurality of first clusters, a first RDS of each of the plurality of first clusters and a number of EDSs belonging to each of the plurality of first clusters, including arranging and displaying the first RDS and the number of EDSs in a format of a plurality of tiles horizontally for the plurality of first clusters.

16. The computer readable medium of claim 14, wherein, The receiving information about a first cluster selected from the plurality of first clusters as a next classification target includes displaying any one of the EDSs belonging to the selected first cluster in a third display area.

17. The computer readable medium of claim 14, wherein, Each of the EDSs includes a vector-valued function of time.

18. The computer readable medium of claim 17, wherein, Each of the EDSs includes a plurality of images varying over time. Each of the EDSs includes a plurality of images varying over time.

Images