Claims:1. A processor implemented method for analyzing sensor signal, comprising:
collecting, via one or more hardware processors, a plurality of sensor signals as input, wherein said plurality of sensor signals belong to a plurality of different sensors;
generating a plurality of generic features from said plurality of sensor signals, via the one or more hardware processors, comprises:
splitting each of the plurality of sensor signals to a plurality of time windows;
transforming each of the plurality of sensor signals in each of the plurality of time windows to time domain, frequency domain, and wavelet domain;
processing the time domain, frequency domain, and wavelet domain based on a generic feature set comprising at least one of or a combination of a plurality of statistics, further wherein the plurality of statistics comprise mean, Standard Deviation, Kurtosis, Skewness, Zero crossings, Energy, Root Mean Square (RMS), Box PierceStatistics, and Hurst Exponent; and
generating a feature matrix corresponding to each of the plurality of sensor signals, wherein the feature matrix provides a unique feature space representation of the plurality of sensor signals;
selecting at least one relevant feature from said feature matrix, by performing a feature selection, via the one or more hardware processors; and
classifying the plurality of sensor signals using a multi-stage binary classification, based on the at least one relevant feature, via the one or more hardware processors.
2. The method as claimed in claim 1, wherein the generic feature set is used to generate a set of consistent features for at least one given sensor signal, by:
collecting at least one sensor signal as a subset of a test data input;
generating the feature matrix for the collected at least one sensor signal, using the generic feature set;
generating the feature matrix for a training data set corresponding to the subset of test data input, using the generic feature set;
computing a matusita distance representing difference in statistical distribution between the feature matrix of the subset of test data input and that of the training data; and
identifying at least one feature that is consistent, in terms of the matusita distance, across the test data and the training data.
3. The method as claimed in claim 2, wherein the at least one feature that is consistent across the test data and the training data is used to handle diverse feature distribution between a source domain and a target domain.
4. The method as claimed in claim 1, wherein said method handles diverse feature distribution between a source domain and a target domain using a misclassification error based weightage mechanism.
5. The method as claimed in claim 1, wherein executing the multi-class classification mechanism comprises:
computing a plurality of cascaded binary classification structures with respect to the at least one selected feature;
generating a generic feature matrix corresponding to each of the cascaded binary classification structures;
recommending at least one feature corresponding to each of the cascaded binary classification structures;
evaluating performance of each of the plurality of cascaded binary classification structures, based on the at least one feature recommended for each of the plurality of cascaded binary classification structures; and
recommending an optimum performing cascaded binary classification structure and corresponding feature set.
6. A system (100), comprising:
a memory module (101) storing instructions;
one or more communication interfaces (103); and
one or more hardware processors (102) coupled to the memory module (101) via the one or more communication interfaces (103), wherein the one or more hardware processors (102) are configured by the instructions to:
collect a plurality of sensor signals as input, wherein said plurality of sensor signals belong to a plurality of different sensors;
generate a plurality of generic features from said plurality of sensor signals, comprises:
splitting each of the plurality of sensor signals to a plurality of time windows;
transforming each of the plurality of sensor signals in each of the plurality of time windows to time domain, frequency domain, and wavelet domain;
processing the time domain, frequency domain, and wavelet domain based on a generic feature set comprising at least one of or a combination of a plurality of statistics, further wherein the plurality of statistics comprise mean, Standard Deviation, Kurtosis, Skewness, Zero crossings, Energy, Root Mean Square (RMS), Box Pierce Statistics, and Hurst Exponent; and
generating a feature matrix corresponding to each of the plurality of sensor signals, wherein the feature matrix provides a unique feature space representation of the plurality of sensor signals;
select at least one relevant feature from said feature matrix, by performing a feature selection; and
classify the plurality of sensor signals using a multi-stage binary classification, based on the at least one relevant feature, via the one or more hardware processors.
7. The system as claimed in claim 6, wherein the instructions when executed by the one or more hardware processors for generating a set of consistent features for at least one given sensor signal using the generic feature set, cause:
collecting the at least one sensor signal as a subset of test data input;
generating the feature matrix for the collected at least one sensor signal, using the generic feature set;
generating the feature matrix for a training data set corresponding to the subset test data input, using the generic feature set;
computing a matusita distance representing difference in statistical distribution between the feature matrix of the subset of test data input and that of the training data; and
identifying at least one feature that is consistent, in terms of the matusita distance, across the test data and the training data.
8. The system as claimed in claim 7, wherein said system is configured to use the at least one feature that is consistent across the test data and the training data to handle diverse feature distribution between a source domain and a target domain.
9. The system as claimed in claim 6, wherein said system handles diverse feature distribution between a source domain and a target domain using a misclassification error based weightage mechanism.
10. The system as claimed in claim 6, wherein the instructions when executed by the one or more hardware processors for executing the multi-class classification cause:
computing a plurality of cascaded binary classification structures with respect to the at least one selected feature;
generating a generic feature matrix corresponding to each of the cascaded binary classification structures;
recommending at least one feature corresponding to each of the cascaded binary classification structures;
evaluating performance of each of the plurality of cascaded binary classification structures, based on the at least one feature recommended for each of the plurality of cascaded binary classification structures; and
recommending an optimum performing cascaded binary classification structure and corresponding feature set. , Description:FORM 2

THE PATENTS ACT, 1970
(39 of 1970)
&
THE PATENT RULES, 2003

COMPLETE SPECIFICATION
(See Section 10 and Rule 13)

Title of invention:

Method and system for sensor signal processing

Applicant

Tata Consultancy Services Limited
A company Incorporated in India under the Companies Act, 1956
Having address:
Nirmal Building, 9th floor,
Nariman point, Mumbai 400021,
Maharashtra, India

The following specification particularly describes the invention and the manner in which it is to be performed.


TECHNICAL FIELD
The disclosure herein generally relates to sensor signal processing, and, more particularly, to a method and system for processing signals from different types of sensors at once for feature extraction and classification.

BACKGROUND
A sensor, by definition, is a device/module that is used for detecting events/changes in its environment and which sends information pertaining to the detected change/event to one or more other electronics devices the sensor is connected to. The sensor signals are then processed and interpreted by the one or more other electronics devices. Different types of sensors are available, each type being suitable for detecting events/changes of a specific type/category. By processing the sensor signals, the electronics device(s) which receives the signal generates/obtains data that may form a part of final output of the electronic device(s), or may be used by the electronic device as an input to generate one or more other inputs.
In a system that collects sensor signal/data from (multiple) various types of sensors, processing of signal/data becomes a cumbersome task. Further, user intervention and inputs maybe required at different stages of data processing. Furthermore, when a system is required to process different types of data (obtained from different types of sensors), the system may fail to collect/fetch right type of information from various types of sensor signals being processed at once.

SUMMARY
Embodiments of the present disclosure present technological improvements as solutions to one or more of the above-mentioned technical problems recognized by the inventors in conventional systems. For example, in one embodiment, a processor implemented method for analyzing sensor signal is provided. The method comprises collecting, via one or more hardware processors, a plurality of sensor signals as input, wherein the plurality of sensor signals belong to a plurality of different sensors. Further, from the plurality of sensor signals, a plurality of generic features are generated via the one or more hardware processors. The feature generation involves: splitting each of the plurality of sensor signals to a plurality of time windows; transforming each of the plurality of sensor signals in each of the plurality of time windows to time domain, frequency domain, and wavelet domain; processing the time domain, frequency domain, and wavelet domain based on a generic feature set comprising at least one of or a combination of a plurality of statistics, further wherein the plurality of statistics comprise mean, Standard Deviation, Kurtosis, Skewness, Zero crossings, Energy, Root Mean Square (RMS), Box Pierce Statistics, and Hurst Exponent; and generating a feature matrix corresponding to each of the plurality of sensor signals, wherein the feature matrix provides a unique feature space representation of the plurality of sensor signals. Further from the feature matrix, at least one relevant feature is selected, by performing a feature selection, via the one or more hardware processors, and the plurality of sensor signals are classified using a multi-stage binary classification, based on the at least one relevant feature, via the one or more hardware processors.
In another aspect, a system (100) is provided. The system 100 comprises of a memory module (101) storing instructions; one or more communication interfaces (103); and one or more hardware processors (102) coupled to the memory module (101) via the one or more communication interfaces (103). The one or more hardware processors (102) are configured by the instructions to collect a plurality of sensor signals as input, wherein the plurality of sensor signals belong to a plurality of different sensors, and then generate a plurality of generic features from the plurality of sensor signals. Generating the plurality of features from the sensor signals comprises of: splitting each of the plurality of sensor signals to a plurality of time windows; transforming each of the plurality of sensor signals in each of the plurality of time windows to time domain, frequency domain, and wavelet domain; processing the time domain, frequency domain, and wavelet domain based on a generic feature set comprising at least one of or a combination of a plurality of statistics, further wherein the plurality of statistics comprise mean, Standard Deviation, Kurtosis, Skewness, Zero crossings, Energy, Root Mean Square (RMS), Box Pierce Statistics, and Hurst Exponent; and generating a feature matrix corresponding to each of the plurality of sensor signals, wherein the feature matrix provides a unique feature space representation of the plurality of sensor signals. Further, at least one relevant feature is selected from the feature matrix, by performing a feature selection, and the plurality of sensor signals are classified using a multi-stage binary classification, based on the at least one relevant feature, via the one or more hardware processors.
In yet another aspect, a non-transitory computer readable medium for analyzing sensor signal is provided. The non-transitory computer readable medium collects, via one or more hardware processors, a plurality of sensor signals as input, wherein the plurality of sensor signals belong to a plurality of different sensors. Further, from the plurality of sensor signals, a plurality of generic features are generated via the one or more hardware processors. The feature generation involves: splitting each of the plurality of sensor signals to a plurality of time windows; transforming each of the plurality of sensor signals in each of the plurality of time windows to time domain, frequency domain, and wavelet domain; processing the time domain, frequency domain, and wavelet domain based on a generic feature set comprising at least one of or a combination of a plurality of statistics, further wherein the plurality of statistics comprise mean, Standard Deviation, Kurtosis, Skewness, Zero crossings, Energy, Root Mean Square (RMS), Box Pierce Statistics, and Hurst Exponent; and generating a feature matrix corresponding to each of the plurality of sensor signals, wherein the feature matrix provides a unique feature space representation of the plurality of sensor signals. Further from the feature matrix, at least one relevant feature is selected, by performing a feature selection, via the one or more hardware processors. Further, the plurality of sensor signals are classified using a multi-stage binary classification, based on the at least one relevant feature, via the one or more hardware processors.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention, as claimed.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings, which are incorporated in and constitute a part of this disclosure, illustrate exemplary embodiments and, together with the description, serve to explain the disclosed principles:
FIG. 1 illustrates an exemplary system for signal processing, according to some embodiments of the present disclosure.
FIG. 2 is a flow diagram depicting steps involved in the process of processing a plurality of signals by the system of FIG. 1, according to some embodiments of the present disclosure.
FIG. 3 illustrates a flow diagram depicting steps involved in the process of generating a consistent feature set by the system of FIG. 1, in accordance with some embodiments of the present disclosure.
FIG. 4 is a flow diagram depicting steps involved in the process of recommending an optimum performing cascaded binary classification structure and corresponding feature set, by the system of FIG. 1, according to some embodiments of the present disclosure.

DETAILED DESCRIPTION OF EMBODIMENTS
Exemplary embodiments are described with reference to the accompanying drawings. In the figures, the left-most digit(s) of a reference number identifies the figure in which the reference number first appears. Wherever convenient, the same reference numbers are used throughout the drawings to refer to the same or like parts. While examples and features of disclosed principles are described herein, modifications, adaptations, and other implementations are possible without departing from the spirit and scope of the disclosed embodiments. It is intended that the following detailed description be considered as exemplary only, with the true scope and spirit being indicated by the following claims.
Referring now to the drawings, and more particularly to FIG. 1 through 4, where similar reference characters denote corresponding features consistently throughout the figures, there are shown preferred embodiments and these embodiments are described in the context of the following exemplary system and/or method.
FIG. 1 illustrates an exemplary system for signal processing, according to some embodiments of the present disclosure. In an embodiment, the system 100 includes one or more hardware processors 102, communication interface(s) or input/output (I/O) interface(s) 103, and one or more data storage devices or memory module 101 operatively coupled to the one or more hardware processors 102. The one or more hardware processors 102 can be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, state machines, graphics controllers, logic circuitries, and/or any devices that manipulate signals based on operational instructions. Among other capabilities, the processor(s) are configured to fetch and execute computer-readable instructions stored in the memory. In an embodiment, the system 100 can be implemented in a variety of computing systems, such as laptop computers, notebooks, hand-held devices, workstations, mainframe computers, servers, a network cloud and the like.
The communication interface(s) 103 can include a variety of software and hardware interfaces, for example, a web interface, a graphical user interface, and the like and can facilitate multiple communications within a wide variety of networks N/W and protocol types, including wired networks, for example, LAN, cable, etc., and wireless networks, such as WLAN, cellular, or satellite. In an embodiment, the communication interface(s) 103 can include one or more ports for connecting a number of devices to one another or to another server.
The memory module(s) 101 may include any computer-readable medium known in the art including, for example, volatile memory, such as static random access memory (SRAM) and dynamic random access memory (DRAM), and/or non-volatile memory, such as read only memory (ROM), erasable programmable ROM, flash memories, hard disks, optical disks, and magnetic tapes. In an embodiment, one or more modules (not shown) of the system 100 can be stored in the memory 101.
FIG. 2 is a flow diagram depicting steps involved in the process of processing a plurality of signals by the system of FIG. 1, according to some embodiments of the present disclosure. The system 100 initially collects (202) one or more sensor signals as input. In an embodiment, the system 100 is configured to collect multiple sensor signals, which may be from same type of sensors or different types of sensors, at once, as input. Further, the sensor signals collected may or may not be from the same domain/class. For example, the system 100 collects physiological signals from one or more users/patients as input. In another example, the plurality of sensor signals may correspond to traffic data collected from a specific geographical area. The traffic sensor data may further have sub-classes based on parameters such as but not limited to type of vehicle/ transport types, route, and so on. For example, assume that in a particular geographic area buses, metro rail, as well as trams are operational, and that the system 100 is collecting traffic data in this particular geographic area using appropriate sensors such as but not limited to accelerometer, and gyro meter. In this scenario, the traffic data may primarily be divided as bus specific data, metro rail specific data, and tram specific data. Sensor signal data from each category may have sub-classes; for example the bus specific data may further include bus numbers, bus routes, and so on. The system 100 is configured to process sensor signals carrying such data, extract appropriate data, and then classify each data as belonging to one or more specific classes. Various steps involved in this process are explained below:
Step 1: Feature extraction and matrix generation (204):
In various embodiments, the system 100 processes the input sensor signals as they are, as well as after splitting the signals to multiple time windows. In various embodiments, the number of windows is dynamically decided by the system 100 based on periodicity of the input signal(s) or may be a fixed number based on a user input. The system 100 may split the signals to windows in the following format:
T_1^'= [t1, t2, t3], T_2^'= [t4, t5, t6],…… T_(N/3)^'= [tN-2, tN-1, tN]
The system 100 then transforms the signals to time domain, frequency domain, and wavelet domain signals (transformed data). The system 100 maintains in an associated reference database, data pertaining to 9 different statistical moments and measures (statistics) which include mean, standard deviation, kurtosis, skewness, zero crossings, energy, root mean square, box pierce statistics, and Hurst exponent, and multiple different combinations of the aforementioned statistics. A few examples of the combinations of statistics are given below:
'Mean of windowedBox-PierceStat2 of DWT (d1)',
'Std of windowedKurtosis of DWT (d2)',
'Variance of amplitude differences of max-min segments',
'Mean of windowedZeroCrossingRate of FFT', and so on.
These statistics and the combinations of statistics form a generic feature set. In an embodiment, more (different types of) statistics may be added to the generic feature set. The system 100 processes the time domain, frequency domain, and the wavelet domain signals using the generic feature set and generates a generic feature matrix (F1, F2, ….FN) corresponding to the input sensor signals, where N is not less than 392. By processing the transformed data and obtaining a representation into the feature space by using the generic feature set, the system 100 extracts features that contain data that provides information pertaining to events/changes being monitored by the sensors. These features then form the generic feature matrix corresponding to the input sensor signal(s).
Step 2: Feature selection (206)
The system 100 then performs a feature selection (the mechanism used for feature selection is covered in Indian patent application 201821019386), by virtue of which the system generates a feature subset comprising at least one selected relevant feature from the generic feature matrix.
Step 3: Classification (208)
The feature subset generated using the feature selection is then processed by the system 100 using a multi-class classification mechanism. By virtue of the classification, the system 100 classifies the signal and the corresponding features as belonging to one or more of a plurality of classes. For example, considering that the sensor signals collected are the signals pertaining to traffic data over a particular area, the system 100 classifies the sensor signals and the features extracted as belonging to tram, buses, and metro rail, and corresponding sub-classes. If the sensor signals collected are physiological signals from one or more patients, then the system 100, by virtue of classification, classifies the signals and corresponding features as matching/indicative of specific health conditions.
In this process, the system 100 initially decides number of classes (C). In various embodiments, the value of ‘C’ maybe selected based on a user input or maybe dynamically learned by the system 100 based on user given annotation or labels provided corresponding to the training data. Further the system 100, based on the number of classes decided, computes a plurality of cascaded binary classification structures corresponding to the feature subset (generated by executing the feature selection). The system 100 further generates, corresponding to each of the plurality of cascaded binary classification structures, corresponding generic feature matrix. The system 100 then recommends a set of features corresponding to each of the cascaded binary classification structures. The system 100 further evaluates performance of each of the plurality of cascaded binary classification structures, based on the user provided cost function (examples F1 score, sensitivity, specificity) as well as using set of feature recommended for each of the plurality of cascaded binary classification structures. Based on the evaluation, the system then recommends an optimum performing cascaded binary classification structure and corresponding feature set for classifying the input signal(s) based on the at least one relevant feature.
In addition to this, the system 100 is configured to handle diverse feature distribution between source and target domains. By virtue of this mechanism, the system 100 identifies features which are consistent across different feature distributions. The system 100, upon receiving a small subset of one or more sensor signals as test data input, generates a feature matrix corresponding to that small subset test data input by executing the aforementioned feature generation mechanism. The system 100 then identifies consistent features of training dataset corresponding to the small sub set of test dataset, and then generates a feature matrix corresponding to the training dataset. Then the system 100 computes a matusita distance which represents difference in statistical distribution between the features metrics corresponding to the small subset of test data input and the training data, using the following equation:
d_M=v(?_(i=1)^d¦?(v(P_i )-v(Q_i ))?^2 ) --- (1)
= v(2-2?_(i=1)^d¦v(P_i Q_i )) ---- (2)
Where,
P and Q are probability density estimates (exemplary case: Gaussian kernel density estimates) two different statistical distributions corresponding to a certain variable.
P_i and Q_i are the probability density estimates corresponding to P and Q, respectively, for the ith value of the variable.
Based on the computed matusita distance, the features are considered as ‘consistent’. For example, if the matusita distance between the training and test features generated from small subset of test data is ‘low’ in comparison with a threshold value, then those features are listed as consistent, which means the features in training and test data are similar in distribution.
Similarly, if the matusita distance between the training and test features is ‘high’ in comparison with a threshold value, then those features are listed as non-consistent (which means the features in training and test data are not similar in distribution).
An alternate approach (misclassification error based weightage mechanism) used by the system 100 to handle diverse feature distribution between source and target domains is explained here. In this approach, a training data (source data) and a small subset of test data (target data) are input to the system 100, wherein the training data and the testing data have different distributions. One or more classifiers of the system 100 are trained on source data, and then by comparing it with the target data, a misclassification error is generated. If the difference in distribution is high, then the misclassification error is high, and in that case, the system 100 assigns a lower weight to the source data. Similarly if the difference in distribution is low, then a larger weightage is assigned to the data. After a few iterations, training data having different distribution have lower weights whereas the training data having same/similar distributions have higher weightage. Instances having higher weightages are used to train classifiers.
The system 100 is also configured to use a metadata based learning approach to improve classifiers used by the system 100. In this approach, the system 100 initially collects different class specific metadata. For example, if physiological data is to be processed, characteristics/features with respect to users (patients), various health conditions (diseases) and so on are collected as the metadata. The collected metadata are then ranked by the system 100, using appropriate mathematical and/or statistical techniques such as but not limited to correlation coefficient, and Pearson Chi-Square. Information pertaining to the metadata, corresponding ranks, and significance are used to train a hierarchical rule engine (not shown in figures) associated with the system 100. As explained in Indian patent application 201821009796, at each stage of the rule engine, classification decision is reached by setting threshold values on the metadata. The threshold values are obtained by applying a similarity measure between different classes. In an embodiment, the similarity is measured using Bhattacharya distance, which gives measure of similarity between two discrete or continuous probability distributions. Steps involved in the process of measuring similarity using Bhattacharya distance is given below:
a) When all subjects having a certain numerical metadata less than or equal to ‘I’ are removed, the remaining subjects in different classes form a frequency histogram
b) A kernel density estimation by using a Gaussian kernel to fit a probability density curve to both histograms is used
c) If the domain, i.e. range of the metadata, is denoted by X, a single value in the domain is denoted by x, and the two probability distributions are denoted by p and q, then Bhattacharyya distance is defined as:
BC (p; q) = P x2X p(x) q(x) --- (3)
The value of BC lies between 0 and 1. More the value of BC, more is the similarity between the two histograms. Hence Bhattacharyya distance can be used as a correlation coefficient (r) between the populations because more a metadata distribution differs in different classes, the better it is for classification.
In the metadata based learning approach, the system 100, upon receiving one or more input values (say, with respect to physiological signals), compares the values with an appropriate rule in the rule engine. The rule engine at each step suggests/recommends action(s) that matches the values of each of the parameters/features. For example, if the input values indicate that age of a patient being monitored is above 45, and if the patient is diabetic, then the chances of the patient having a Coronary Artery Disease (CAD) is more. If at least one of the aforementioned conditions/statements is false, then the system 100 asks one or more other questions with respect to rules defined in the rule engine, and accordingly performs health condition assessment of the user (classifies as CAD or non-CAD).
FIG. 3 illustrates a flow diagram depicting steps involved in the process of generating a consistent feature set by the system of FIG. 1, in accordance with some embodiments of the present disclosure. Generating a consistent set of features is done so as to identify one or more features that are consistent across a source and target data having diverse distributions. The system 100 initially collects (302) one or more sensor signals as a subset of test data input, and then generates (304) a feature matrix corresponding to that subset of test data input by executing the feature generation mechanism. The system 100 then identifies a training dataset corresponding to the small subset of test dataset, and then generates (306) a feature matrix corresponding to the training dataset. Then the system 100 computes (308) a matusita distance which represents difference in statistical distribution between the features metrics corresponding to that test data input and the training data, using equations (1) and (2).
Based on the computed matusita distance, the features are classified/identified (310) as ‘consistent’. For example, if the matusita distance between the training and test features is ‘low’ in comparison with a threshold value, then the features are listed as consistent (which means the features in training and test data are similar in distribution). Similarly, if the matusita distance between the training and test features is ‘high’ in comparison with a threshold value (example threshold value = 1), then the features are identified as not consistent (which means the features in training and test data are not similar in distribution). In various embodiments, the steps in method 300 may be performed in the same order or in a different order, and one or more of the steps may be skipped if required.
FIG. 4 is a flow diagram depicting steps involved in the process of recommending an optimum performing cascaded binary classification structure and corresponding feature set, by the system of FIG. 1, according to some embodiments of the present disclosure. In this process, the system 100 processes the feature subset (generated by executing the feature selection on the generic feature matrix corresponding to the input sensor signals) and computes (404) a plurality of cascaded binary classification structures. The system 100 then recommends (406) at least one feature corresponding to each of the cascaded binary classification structures. The system 100 further evaluates (408) performance of each of the plurality of cascaded binary classification structures, based on the at least one feature recommended for each of the plurality of cascaded binary classification structures. Based on the evaluation, the system then recommends (410) an optimum performing cascaded binary classification structure and corresponding feature set. In various embodiments, the steps in method 400 may be performed in the same order or in a different order, and one or more of the steps may be skipped if required.
The written description describes the subject matter herein to enable any person skilled in the art to make and use the embodiments. The scope of the subject matter embodiments is defined by the claims and may include other modifications that occur to those skilled in the art. Such other modifications are intended to be within the scope of the claims if they have similar elements that do not differ from the literal language of the claims or if they include equivalent elements with insubstantial differences from the literal language of the claims.
The embodiments of present disclosure herein addresses unresolved problem of processing of sensor signals of different types together, for feature extraction and for classifying features extracted from sensor signals. The embodiment, thus provides a generic feature set for generating a generic feature matrix from an input sensor signal. Moreover, the embodiments herein further provides a multi-class binary classification mechanism to classify features generated from an input sensor signal as belonging to one or more of specific classes.
It is to be understood that the scope of the protection is extended to such a program and in addition to a computer-readable means having a message therein; such computer-readable storage means contain program-code means for implementation of one or more steps of the method, when the program runs on a server or mobile device or any suitable programmable device. The hardware device can be any kind of device which can be programmed including e.g. any kind of computer like a server or a personal computer, or the like, or any combination thereof. The device may also include means which could be e.g. hardware means like e.g. an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a combination of hardware and software means, e.g. an ASIC and an FPGA, or at least one microprocessor and at least one memory with software modules located therein. Thus, the means can include both hardware means and software means. The method embodiments described herein could be implemented in hardware and software. The device may also include software means. Alternatively, the embodiments may be implemented on different hardware devices, e.g. using a plurality of CPUs.
The embodiments herein can comprise hardware and software elements. The embodiments that are implemented in software include but are not limited to, firmware, resident software, microcode, etc. The functions performed by various modules described herein may be implemented in other modules or combinations of other modules. For the purposes of this description, a computer-usable or computer readable medium can be any apparatus that can comprise, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.
The illustrated steps are set out to explain the exemplary embodiments shown, and it should be anticipated that ongoing technological development will change the manner in which particular functions are performed. These examples are presented herein for purposes of illustration, and not limitation. Further, the boundaries of the functional building blocks have been arbitrarily defined herein for the convenience of the description. Alternative boundaries can be defined so long as the specified functions and relationships thereof are appropriately performed. Alternatives (including equivalents, extensions, variations, deviations, etc., of those described herein) will be apparent to persons skilled in the relevant art(s) based on the teachings contained herein. Such alternatives fall within the scope and spirit of the disclosed embodiments. Also, the words “comprising,” “having,” “containing,” and “including,” and other similar forms are intended to be equivalent in meaning and be open ended in that an item or items following any one of these words is not meant to be an exhaustive listing of such item or items, or meant to be limited to only the listed item or items. It must also be noted that as used herein and in the appended claims, the singular forms “a,” “an,” and “the” include plural references unless the context clearly dictates otherwise.
Furthermore, one or more computer-readable storage media may be utilized in implementing embodiments consistent with the present disclosure. A computer-readable storage medium refers to any type of physical memory on which information or data readable by a processor may be stored. Thus, a computer-readable storage medium may store instructions for execution by one or more processors, including instructions for causing the processor(s) to perform steps or stages consistent with the embodiments described herein. The term “computer-readable medium” should be understood to include tangible items and exclude carrier waves and transient signals, i.e., be non-transitory. Examples include random access memory (RAM), read-only memory (ROM), volatile memory, nonvolatile memory, hard drives, CD ROMs, DVDs, flash drives, disks, and any other known physical storage media.
It is intended that the disclosure and examples be considered as exemplary only, with a true scope and spirit of disclosed embodiments being indicated by the following claims.