DESC:The following specification particularly describes and ascertains the nature of this invention and the manner in which it is to be performed:-

CROSS REFERENCE TO RELATED APPLICATIONS
This application is based on and derives the benefit of Indian Provisional Applications 1981/CHE/2015, the contents of which are incorporated herein by reference.

TECHNICAL FIELD
[001] The embodiments herein generally relate to the field of data analytics and more particularly to analytics of sensor data received from a plurality of sensors.

BACKGROUND
[002] Sensor data analytics is a rapidly developing field and involves applying analytics to raw sensor data to facilitate detection of at least one event of interest. Events of interest may be an anomaly or any pre-defined event. As use of devices, which can be a part of Internet of Things (IoT) increases, sensor data analytics can contribute to be a key component in driving new applications by providing new benefits to consumers. Development of Micro Electro Mechanical Systems (MEMS) enables miniaturization, mass production, and cost reduction of many sensors. In particular, MEMS inertial sensors such as acceleration sensors, angular velocity sensors (such as gyroscopes) and so on in an Inertial Measurement Unit (IMU) and the like are getting popular with portable or wearable electronic devices such as smart phones, tablets, hand gears and widely used for sensor data analytics.
[003] For applications based on sensor data analytics, it is critical to detect the relevant part of a signal captured by a sensor and identify different events of interest from the signal with an accurate estimation of event location/time-point and event occurrence window. Most existing off-the-shelf event detection techniques are unable to detect events when Signal-to-Noise Ratio (SNR) is considerably low. For example, in mobile sensor based car analytics engine noise and road disturbances are a predominant source of noise when sensor data is acquired from the inertial sensors of a mobile device. Detecting sharp driving maneuvers is relatively easier in presence of the background noise as such events generate patterns with high SNR; whereas, normal maneuvers such as overtaking or incomplete lane change are difficult to detect due to very low SNR. Further, existing methods are based on pattern matching, wherein sensor data corresponding to a pattern identified in the signal captured by one or more sensors is matched with data corresponding to a stored pattern to detect the event of interest. However, such pattern matching requires maintaining a library of known patterns, which is a static technique limiting usage for dynamic events in real time. Further, such existing pattern matching may be a challenge in high noise environment and when data to be analyzed for event detection is a multi-dimensional data derived from the signals captured by plurality of sensors of the electronic device.

OBJECT OF INVENTION
[004] The principal object of the embodiments herein is to provide methods and systems for detecting occurrence of one or more events by analyzing data, received from a plurality of sensors at pre-defined intervals, using a Hidden Morkov Model (HMM).

SUMMARY
[005] In view of the foregoing, an embodiment herein provides a method for detecting occurrence of at least one event by analyzing data received from a plurality of sensors at pre-defined intervals. The method comprises pre-processing the data received from the plurality of sensors to generate a pre-processed data. Further, the method comprises converting the pre-processed data from a multi-dimensional data to a linear data by computing slopes for the preprocessed data corresponding to each pre-defined interval. Furthermore, the method comprises detecting occurrence of the at least one event among a plurality of preset events of the Hidden Markov Model (HMM) by analyzing the computed slopes.
[006] Embodiments further disclose an electronic device for detecting occurrence of at least one event by analyzing data received from a plurality of sensors at pre-defined intervals is described. The electronic device comprises an event detection module configured to pre-process the data received from the plurality of sensors to generate a pre-processed data. Further, the event detection module is configured to convert the pre-processed data from a multi-dimensional data to a linear data by computing slopes for the preprocessed data corresponding to each pre-defined interval. Furthermore, the event detection module is configured to detect occurrence of the at least one event among a plurality of preset events of the Hidden Markov Model (HMM) by analyzing the computed slopes.
[007] These and other aspects of the embodiments herein will be better appreciated and understood when considered in conjunction with the following description and the accompanying drawings. It should be understood, however, that the following descriptions, while indicating embodiments and numerous specific details thereof, are given by way of illustration and not of limitation. Many changes and modifications may be made within the scope of the embodiments herein without departing from the spirit thereof, and the embodiments herein include all such modifications.

BRIEF DESCRIPTION OF FIGURES
[008] The embodiments of this invention are illustrated in the accompanying drawings, throughout which like reference letters indicate corresponding parts in the various figures. The embodiments herein will be better understood from the following description with reference to the drawings, in which:
[009] FIG. 1 illustrates a plurality of components of an electronic device, according to embodiments as disclosed herein;
[0010] FIG. 2 is a flow diagram illustrating a method for detecting occurrence of one or more events by analyzing data received from a plurality of sensors at pre-defined intervals using a Hidden Morkov Model (HMM), according to embodiments as disclosed herein;
[0011] FIG. 3 is a flow diagram illustrating method for detecting occurrence of one or more events among a plurality of preset events of the HMM by identifying current state based on probabilities computed for the data using slopes representing pre-processed data, according to embodiments as disclosed herein; and
[0012] FIG. 4 is an example illustrating detecting occurrence of events by analyzing received signal from plurality of sensors for driving analytics during a driving maneuver, according to embodiments as disclosed herein.

DETAILED DESCRIPTION
[0013] The embodiments herein and the various features and advantageous details thereof are explained more fully with reference to the non-limiting embodiments that are illustrated in the accompanying drawings and detailed in the following description. Descriptions of well-known components and processing techniques are omitted so as to not unnecessarily obscure the embodiments herein. The examples used herein are intended merely to facilitate an understanding of ways in which the embodiments herein may be practiced and to further enable those of skill in the art to practice the embodiments herein. Accordingly, the examples should not be construed as limiting the scope of the embodiments herein.
[0014] The embodiments herein achieve methods and systems for detecting occurrence of one or more events by analyzing data received by an electronic device from a plurality of sensors at pre-defined intervals using a Hidden Morkov Model (HMM). Different states of interest can be defined for the HMM based on one or more preset events that include one or more events of interest. Further, any event other than the event of interest is identified as a non-event in the preset events and is also one state among plurality of states defined for the HMM. The method includes pre-processing the data and analyzing the pre-processed data using HMM. The pre-processed data is generated by pre-processing the data received from plurality of sensors using one or more filters. Further, the preprocessed data, which time series data is a multidimensional data and is converted to a linear data by computing slopes of the data being analyzed. Thus, the slopes represent the data in single dimension. The HMM identifies a current state for the pre-processed data based on probabilities computed for the data using slopes representing the pre-processed data. Thus, if the current state is a state of interest as defined by one or more preset events, then the HMM compares the probabilities of the data being analyzed to identify an event window for the identified state based on whether the computed probability is above or below a predefined threshold for the particular identified event. The HMM, based on states and probability computation enables identification of the event from the data being monitored even during low Signal to Noise Ratio (SNR) scenarios.
[0015] In an embodiment, the electronic device is a smart phone, a tablet, a personal digital assistant, a wearable device and the like.
[0016] In an embodiment, the plurality of sensors include Micro Electro Mechanical Systems (MEMS) inertial sensors such as acceleration sensors, angular velocity sensors (such as gyroscopes) and the like that may be within an Inertial Measurement Unit (IMU). In an embodiment, the data may be received from other sensors such as environmental sensors, cameras, location sensors such as Global Positioning System (GPS) and the like.
[0017] Referring now to the drawings, and more particularly to FIGS. 1 through 4, where similar reference characters denote corresponding features consistently throughout the figures, there are shown embodiments.
[0018] FIG. 1 illustrates a plurality of components of an electronic device 100, according to embodiments as disclosed herein.
[0019] Referring to figure 1, the electronic device 100 is illustrated in accordance with an embodiment of the present subject matter. In an embodiment, the electronic device 100 may include at least one processor 102, an input/output (I/O) interface 104 (herein a configurable user interface), a memory 106. The at least one processor 102 may be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, state machines, logic circuitries, and/or any devices such as a mobile phone, a laptop, a personal digital assistant, a tablet, a wearable device and so on that manipulate signals based on operational instructions. Among other capabilities, the at least one processor 102 is configured to fetch and execute computer-readable instructions stored in the memory 106.
[0020] The I/O interface104 may include a variety of software and hardware interfaces, for example, a web interface, a graphical user interface and the like.
[0021] In an embodiment the electronic device 100 may only have an input interface such as an On-board Diagnostics (OBD) –II dongle.
[0022] In an embodiment, the electronic device 100 may include a sensor unit 114 that may include plurality of sensors such as the MEMS inertial sensors providing the acceleration sensors, the angular velocity sensors (such as gyroscopes) and the like. The MEMS inertial sensor may be included in the sensor unit as the IMU. In an embodiment, the sensor unit 114 may include one or more external sensor such as a camera and the like that provide the data corresponding to one or more environmental parameters being monitored. The I/O interface 104 may allow the electronic device 100 to communicate with any external sensors and with other devices such as wearable devices and other electronic devices. The I/O interface 104 may facilitate multiple communications within a wide variety of networks and protocol types, including wired networks, for example, Local Area network (LAN), cable and the like and wireless networks, such as Wireless LAN, cellular, Device to Device (D2D) communication network, Wi-Fi networks and so on. The modules 108 include routines, programs, objects, components, data structures, and so on, which perform particular tasks, functions or implement particular abstract data types. The modules 108 may include programs or coded instructions that supplement applications and functions of the electronic device 100 such as functions of an event detection module 112. Data 110, amongst other things, serves as a repository for storing data processed, received, and generated by one or more of the modules 108 and/or the event detection module 112.
[0023] In an embodiment, the event detection module 112 can be configured to receive data from a plurality of sensors of the sensor unit 114 at pre-defined intervals. For example, the predefined interval can be in the range of milliseconds. The data received from the sensors such as the accelerometer of the IMU is a time variant data providing changes in the measured parameter such as velocity of a subject or object being monitored. Further, the event detection module 112 can be configured to pre-process data received from the plurality of sensors to generate the pre-processed data. The pre-processing of sensor signals is performed to provide signal conditioning and reduce noise and improve the SNR of the data received from the plurality of sensors using one or more filters. The filter may be a combination filter such as a Low Pass Filter and a Kalman filter or the like. Since the data is received is time series data, represented by a multidimensional vector, the event detection module 112 can be configured to convert the pre-processed data from multi-dimensional data to linear data, which reduces the mathematical complexity for further processing. The pre-processed data is converted to the linear data by computing slopes of the pre-processed data being analyzed. Thus, the computed slopes represent the data in single dimension and a single dimension vector. Thus, the electronic device enables significant reduction in dimension of the observation vector (multi dimensional data) without losing local dependency and overall signal pattern. To reduce further mathematical complexities the slope values are adjusted to absolute values within a pre-defined range. For example, the slope is adjusted between values 0 – 4. All slope values above 4 are considered to be positive.
[0024] Further, the event detection module 112 can be configured to detect occurrence of one or more events among the plurality of preset events defined in the HMM. The event detection module 112 can be configured to compute probabilities for the data for each interval among the pre-defined intervals using the HMM and the computed slopes. The probabilities are computed for every preset event defined in the HMM. The HMM is a finite state standard stochastic model, defined by at least one parameter, wherein the at least on parameter comprises a transition probability matrix and an Emission probability matrix. Further, the preset events comprises one or more events of interest and a non-event, wherein the non-event indicates that none of the event among the preset event has occurred and is been detected. Thus, when x, y, z are the preset events in the HMM, non occurrence of any of the x, y and z event is termed as the non-event and each of the four events x. y, z and the non event is assigned a unique state. To analyze the preprocessed data for detecting occurrence of one or more events, the HMM is trained offline. The offline training of the HMM model comprises an intuitive trial solution to roughly define each of the plurality of states (event and no Event) and optimizing the model using an Expectation Maximization (Baum-Welch) Algorithm. The HMM of the event detection module 112 receives the slopes as an input parameter and can be configured to compute probabilities for the plurality of preset events using the slopes. The HMM can be configured to receive states of interest that can be defined by a user for the HMM. The states of the HMM are defined for one or more preset events that include one or more events of interest and a non event (any event other that the events of interest. Further, any event other than the event of interest is identified as a non-event in the preset events and is also one state among plurality of states defined for the HMM. For example the preset events in a driving analytics application can be change of lane for overtaking, a mere change of lane form low speed lane to higher speed lane and the like. The method includes pre-processing the data and analyzing the pre-processed data using HMM. The pre-processed data is generated by pre-processing the data received from plurality of sensors using one or more filters. The filters can be a combination filter such as a low pass filter and a Kalman filter and the like. Further, the preprocessed data, which is a time series data is a multidimensional data and is converted to a linear data by computing slopes of the data being analyzed. Thus, the slopes represent the data in single dimension. The HMM identifies a current state, from among the defined states corresponding to the pre-set events, for the pre-processed data based on probabilities computed for the data using slopes representing the pre-processed data. For example, the current state can be a state corresponding to an event among the preset events for whom the computed probability is maximum. Thus, if the current state from among the state of interest (corresponds to a event of interest from the events of interest preset) then the HMM compares the computed probabilities of the data being analyzed to identify an event window for the identified state based on whether the computed probability is above or below a predefined threshold for the particular identified event corresponding to the stat..
[0025] In an embodiment, functions of the event detection module 112 can split. The major functions such as analyzing the pre-processed data using HMM that require higher processing capabilities can be carried out in a cloud server, whereas the other functions such as pre-processing that may comparatively require low computation capabilities can be performed on the electronic device 100.
[0026] The names of the components and modules of the electronic device 100 are illustrative and need not be construed as a limitation.
[0027] FIG. 2 is a flow diagram illustrating a method 200 for detecting occurrence of one or more events from data received from the plurality of sensors at pre-defined intervals using the HMM, according to embodiments as disclosed herein.
[0028] At step 202, the method 200 includes allowing the event detection module 112 of the electronic device 100 to receive data from the plurality of sensors of the sensor unit 114 at pre-defined intervals. At step 204, the method 200 includes allowing the event detection module 112 to pre-process data received from the plurality of sensors to generate the pre-processed data. For example, the predefined interval can be in the range of milliseconds. The pre-processing of sensor signals is performed to provide signal conditioning and reduce noise and improve the SNR of the data received from the plurality of sensors using one or more filters. The filter may be a combination filter such as a Low Pass Filter and a Kalman filter or the like. At step 206, the method 200 includes allowing the event detection module 112 to convert the pre-processed data from a multi-dimensional data to the linear data. The pre-processed data is converted to the linear data by computing slopes of the pre-processed data being analyzed. Thus, the computed slopes represent the data in single dimension and the single dimension vector.
[0029] At step 208, the method 200 includes allowing the event detection module 112 to detect occurrence of one or more events among the plurality of preset events defined in the HMM. The event detection module 112 can be configured detect occurrence of one or more events among the plurality of preset events of the Hidden Markov Model (HMM) by analyzing the data using the HMM. The analysis of the pre-processed data using the HMM is explained in conjunction with FIG. 3.
[0030] A use example explained in conjunction with FIG. 4 describes how the method 200 detects the event.
[0031] For a fixed class of problem such as deriving maneuvering, hand gesture, racket swing, and so on, the embodiments herein perform one time estimation of HMM parameters to detect events from multiple test data. Regular updating of model parameters, in most cases, is not necessary unless there is a complete regime change in the subject being monitored.
[0032] Modeling framework and subsequent outputs are flexible and amenable for further applications in embodiments herein, such as driver profiling, fuel consumption modeling for car analytics, player profiling for sports analytics, and so on.
[0033] The various actions in method 200 may be performed in the order presented, in a different order or simultaneously. Further, in some embodiments, some actions listed in FIG. 2 may be omitted.
[0034] FIG. 3 is a flow diagram illustrating method 300 for detecting occurrence of one or more events among a plurality of preset events of the HMM by identifying current state based on probabilities computed for the data using slopes representing the pre-processed data, according to embodiments as disclosed herein.
[0035] At step 302, the method 300 includes allowing the HMM of the event detection module 112 to receive states defined for the HMM. At step 304, the method 300 includes allowing the HMM of the event detection module to compute probabilities using slopes corresponding to the pre-processed data. At step 306, the method 300 includes allowing event detection module to identify a current state based on the computed probabilities for each of the preset events. For example, the current state can be a state corresponding to an event among the preset events for whom the computed probability is maximum. At step 308, the method 300 includes allowing the HMM of event detection module to detect the event if the computed probabilities are above a pre-defined probability threshold corresponding to the event. The HMM marks an event window for the current state, wherein the probabilities within the event window are above the predefined threshold. The event window is identified as a portion of the data being analyzed where the probability computed is above the predefined probability threshold for the event.
[0036] The various actions in method 300 may be performed in the order presented, in a different order or simultaneously. Further, in some embodiments, some actions listed in FIG. 3 may be omitted.
[0037] FIG. 4 is an example illustrating detecting occurrence of events by analyzing received signal from plurality of sensors for driving analytics during a driving maneuver, according to embodiments as disclosed herein. The data is received from plurality of sensors of the IMU in the smart phone. The preset events of interest can be to detect lateral maneuvers such as lane changes and turns. For such events, SNR is generally very low due to engine vibrations, road conditions and steering micro-corrections. Existing or off-the-shelf event detection methods generally fail to either detect events or to make things worse, inappropriately chops or clubs events’ window. Embodiments herein are able to detect events with accurate event window from Gyroscope Z-axis signal (also velocity) after pre-processing using Low Pass and Kalman Filter. Fixing a hard probability threshold of 0.9, normal maneuvering events are detected with accurate event window as indicated by the dotted rectangles. Thus probability computed for an event 402 is high (approximately 1) (408) within the event window and zero outside the event window. Further probabilities for events other than the event 402 (identified as non-event) are low and below a probability threshold 406 within the event window. Less typical events such as overtaking, which is typically accelerating amidst a lane change and then an immediate lane change/continuing in the new lane and so on are also detected. Based on estimated model parameters, applications related to fuel consumption modeling, driver profiling (in particular, driver aggressiveness while cruising and maneuvering), and the like can be developed.
[0038] The embodiments disclosed herein can be implemented through at least one software program running on at least one hardware device and performing network management functions to control the network elements. The network elements shown in FIG. 1 through FIG. 4 include blocks which can be at least one of a hardware device, or a combination of hardware device and software module.
[0039] The foregoing description of the specific embodiments will so fully reveal the general nature of the embodiments herein that others can, by applying current knowledge, readily modify and/or adapt for various applications such specific embodiments without departing from the generic concept, and, therefore, such adaptations and modifications should and are intended to be comprehended within the meaning and range of equivalents of the disclosed embodiments. It is to be understood that the phraseology or terminology employed herein is for the purpose of description and not of limitation. Therefore, while the embodiments herein have been described in terms of preferred embodiments, those skilled in the art will recognize that the embodiments herein can be practiced with modification within the spirit and scope of the embodiments as described herein.
,CLAIMS:STATEMENT OF CLAIMS
We claim:
1. A method for detecting occurrence of at least one event from data received from a plurality of sensors at pre-defined intervals, the method comprising:
pre-processing, by an electronic device, the data received from the plurality of sensors to generate a pre-processed data;
converting, by the electronic device, the pre-processed data from a multi-dimensional data to a linear data by computing slopes for the pre-processed data corresponding to each pre-defined interval; and
detecting, by the electronic device, occurrence of the at least one event among a plurality of preset events of the Hidden Markov Model (HMM) by analyzing the slopes for the pre-processed data using the HMM.
2. The method as claimed in claim 1, wherein the pre-processing comprises filtering the data using at least one filter.
3. The method as claimed in claim 1, wherein values of the slopes, so computed, are adjusted to absolute values within a pre-defined range.
4. The method as claimed in claim 1, wherein detecting occurrence of the event among the plurality of preset events of the HMM comprises:
computing probabilities for the plurality of preset events using the slopes and the HMM, wherein each event among the plurality of events is assigned a unique state;
identify a current state from among plurality of states based on the computed probabilities;
detect the event if the computed probabilities are above a pre-defined probability threshold corresponding to the event; and
mark an event window for the current state, wherein the probabilities within the event window are above the predefined threshold.
5. The method as claimed in claim 4, wherein the HMM model is trained off-line to identify the plurality of preset events defined by a user and assign the plurality of preset events with the plurality of states.
6. A electronic device for detecting occurrence of at least one event from data received from a plurality of sensors at pre-defined intervals, wherein the electronic device comprises a event detection module configured to:
pre-process the data received from the plurality of sensors to generate a pre-processed data;
convert the pre-processed data from a multi-dimensional data to a linear data by computing slopes for the pre-processed data corresponding to each pre-defined interval; and
detect occurrence of the at least one event among a plurality of preset events of the Hidden Markov Model (HMM) by analyzing the slopes for the pre-processed data using the HMM.
7. The electronic device as claimed in claim 6, wherein the event detection module is configured to perform pre-processing of the data by filtering the data using at least one filter.
8. The electronic device as claimed in claim 6, wherein values of the slopes, so computed, are adjusted to absolute values within a pre-defined range.
9. The electronic device as claimed in claim 6, wherein the event detection module is configured to detect occurrence of the event among the plurality of preset events of the HMM by:
computing probabilities for the plurality of preset events using the slopes and the HMM, wherein each of the at least one event among the plurality of events is assigned a unique state;
identify a current state from among plurality of states based on the computed probabilities;
detect the at least one event when the computed probabilities are above a pre-defined probability threshold corresponding to the at least one event; and
mark an event window corresponding to the current state, wherein the probabilities within the event window are above the predefined threshold.
10. The electronic device as claimed in claim 9, wherein the HMM model is trained off-line to identify the plurality of preset events defined and assign the plurality of preset events with the plurality of states.

Dated this 18th of April 2016 Signature:
Name of the Signatory: Dr. Kalyan Chakravarthy