WO2014191805A1

WO2014191805A1 - Devices and methods for determination of cognitive load

Info

Publication number: WO2014191805A1
Application number: PCT/IB2014/000792
Authority: WO
Inventors: Arijit Sinharay; Debatri CHATTERJEE; Amit Konar
Original assignee: Tata Consultancy Services Limited
Priority date: 2013-05-27
Filing date: 2014-05-21
Publication date: 2014-12-04

Abstract

The present subject matter discloses device(s) and method(s) for determination of cognitive load on subjects. According to the present subject matter, the device(s) implement the described method(s), where the method(s) include, providing a logical task to a subject, where the logical task is based on at least one propositional statement, and is from predefined logical tasks of predefined complexity levels. Electroencephalography (EEG) signals are obtained from at least one brain lobe of the subject, while the subject is executing the logical task. A predefined set of EEG features are evaluated from the obtained EEG signals and the cognitive load of the subject is classified based on the predefined set of EEG features.

Description

DEVICES AND METHODS FOR DETERMINATION OF COGNITIVE LOAD

TECHNICAL FIELD

[0001] The present subject matter relates, in general, to determination of cognitive load on subjects and, particularly but not exclusively, to determination of cognitive load through Electroencephalography (EEG) on subjects.

BACKGROUND

[0002] In the increasingly competitive world, the ability to perform multitasking has become exceedingly important for effective time management and effective productivity at workplaces or home. The multi-tasking and the performance depend on cognition and cognitive abilities of an individual. Cognition refers to the mental faculties of a person engaged in cognitive tasks related to reasoning, learning, planning, memorizing, problem solving, decision making, and such. The psychological load exerted on the individuals, while execution of such cognitive tasks, is referred to as cognitive load.

[0003] In order to assess cognitive ability of an individual, and determine the associated cognitive load, brain functioning of the individual during execution of the cognitive tasks is decoded by Brain Signal and Image Processing (BSIP) equipment. The BSIP equipment includes an Electroencephalogram (EEG) device, a functional magnetic resonance imaging (fMRI) machine, a functional near infrared spectroscope (fNIRs), an Electrocorticograph (ECoG), a Positron Emission Tomography (PET) machine. However, EEG-based equipment is generally used for its capability to determine a direct measure of brain activity, and also for its inexpensiveness, exceptional temporal resolution and non-invasive characteristics.

BRIEF DESCRIPTION OF DRAWINGS

[0004] The detailed description is described with reference to the accompanying figures. In the figures, the left-most digit(s) of a reference number identifies the figure in which the reference number first appears. The same numbers are used throughout the drawings to reference like features and components.

[0005] Fig. 1 illustrates a system environment with a cognitive load determination device, in accordance with an implementation of the present subject matter.

[0006] Fig. 2 illustrates a method to determine cognitive load on a subject, in accordance with an implementation of the present subject matter.

[0007] It should be appreciated by those skilled in the art that any block diagrams herein represent conceptual views of illustrative devices embodying the principles of the present subject matter. Similarly, it will be appreciated that any flow charts, flow diagrams, state transition diagrams, pseudo code, and the like represent various processes which may be substantially represented in computer readable medium and so executed by a computing device or processor, whether or not such computing device or processor is explicitly shown.

DETAILED DESCRIPTION

[0008] Method(s) and device(s) to determine cognitive load through

Electroencephalography (EEG) on subjects, while the subjects are executing logic- based tasks, are described.

[0009] Cognitive load, although a fuzzy concept, has been a useful measure for cognitive psychologists, educators, managers at workplaces, etc., for estimation of cognition on individuals. Cognitive load refers to the total amount of mental activity imposed on working memory of an individual at any instance in time. In general, cognitive loads are of three types, namely intrinsic cognitive load, extraneous cognitive load, and germane cognitive load. The intrinsic cognitive load refers to a cognitive load due to the complexity of a task. For example, drawing a square is easier than drawing a portrait. Therefore, drawing a portrait imparts a higher ognitive load on an individual than that by drawing a square. The extraneous type refers to a degree of cognitive load depending on method of representation of information. For example, a graphical representation of information imparts a lower cognitive load as compared to that by a tabular representation of information. The germane type refers to a cognitive load generated during learning of a new rule or a new technique. The determination of cognitive load encompasses several real world applications like, estimating the performance and abilities of an individual at a work place before assigning him a task, understanding the health status of an individual, etc.

[0010] Cognitive load is generally determined based on brain activity of the individual, while the individual is executing a cognitive task. Decoding of the brain activity of the individual is achieved generally by a brain-computer interface (BCI) system that detects changes in electrical signals in specific sections of the brain, while the individual is performing the cognitive tasks.

[0011] Methods for determination of cognitive load based on performance of memory-based tasks, such as memorizing a set of words or a sequence of numbers, are generally known. Such methods determine the cognitive load based on the extent of memorizing capabilities of the individual. Conventional methods do not take into consideration the logical reasoning, decision-making abilities and analytical abilities of the individual, but rather rely on the memorizing capabilities of the individual, and thus may not provide a substantially accurate measure of the cognitive load.

[0012] Conventional methods also include inferring cognitive load of the individual based on the individual's performance or behavior. In such methods, the time taken for completion of a particular task is correlated to the cognitive load handling abilities of an individual. For example, the cognitive load may be inferred based on the speed with which the individual reads a document. Further, some conventional methods determine cognitive load based on subjective measures, such as self-rating scales. For example, intrinsic difficulty to read a document from a particular domain or expertise by the individual working in a particular domain forms the basis of inferring the cognitive load. Such methods are not capable of determining cognitive load without having prior information about the individual in that particular domain. Further, these methods do not consider psychological or emotional state and logical-reasoning abilities of the individual. Thus, the cognitive load cannot be determined substantially accurately.

[0013] Existing physiological modalities of cognitive load determination measure the manifestation of the cognitive task on the organs or body parts of the subject. The well-known modalities in this regard include: pupil dilatation, heart rate and changes in galvanic skin resistance. These methods consider submitting a set of stimuli to the individual and monitoring the physiological responses to the provided set of stimuli, such as monitoring change in pupil dilatation, heart rate or galvanic skin responses. The degree of change associated with such bodily functions of an individual, in response to the stimuli, is indicative of the degree of cognitive load on the individual. However, such methods do not consider the cognitive load shared by the brain, and so ignores individual's psychological or emotional state and logical - reasoning abilities. Thus, the cognitive load cannot be determined substantially accurately.

[0014] The present subject matter describes devices and methods to determine cognitive load through Electroencephalography (EEG) on subjects while the subjects are executing logic-based tasks. The subjects include individuals performing logic-based tasks. In an implementation, the subjects may include software programmers involved in logical analysis of program codes. The logic-based tasks, also referred to as logical tasks, to be solved by the subjects may be based on one or more propositional statements, and the subjects may have to execute the propositional statements in order to solve a logical task. Generally, propositional statements have two possible truth values: true and false. For example, let us consider a logical task in which the characteristic of Alice is to be determined. The logical task includes a fact that "Alice is smart" and a propositional statement that "Alice is honest if Alice is smart." The described example has one "If-Then" type propositional statement incorporated in the logical task. Based on these details, the individual performing the tasks can conclude that the Alice is both honest and smart.

[0015] Further, logical tasks based on propositional statements are associated with predefined complexity levels, such as low, medium, and high. The complexity may be ascertained based on the number of propositional statements, incorporated in the logical task. For example, the logical task mentioned above has low complexity level as the number of propositional statements involved is one.

[0016] In an implementation of the present subject matter, for determining the cognitive loads of subjects, the subjects under examination are presented with logical tasks. The logical tasks are based on one or more propositional statements, and are distributed across a number of predefined complexity levels. While the subjects are performing the logical tasks, EEG signals are obtained from brain lobes of each of the subjects. The EEG measurements allow for detection of changes in electrical signals in the brain when the subject is presented with and performing the logical tasks. The EEG signals may be acquired from the brain lobes responsible for logical thinking, for example, the frontal brain lobe. Thus, the EEG signals are indicative of the electrical changes in the brain when the subjects are performing logical tasks.

[0017] Further, based on the EEG signals, a predefined set of EEG features are evaluated from the EEG signals of each of the subjects. The predefined set of EEG features may include at least one of a log variance, a Total Band Power (TBP), a Log Band Power Ratio (LBPR), and Hjorth parameters. In an implementation, the predefined set of EEG features can be determined by presenting the same set of predefined logical tasks with the predefined complexity levels to multiple test- subjects, acquire EEG signals from the test-subjects, evaluate a plurality of EEG features responsible for logical thinking from the EEG signals of the test-subjects, and further identifying common discriminative EEG features from the plurality of EEG features for the test-subjects. A common discriminative EEG feature is an EEG feature identified based on distribution of that EEG feature into classes associated with predefined complexity levels across the test-subjects. The common discriminative EEG features are considered as the predefined set of EEG features.

[0018] Further, based on the evaluated set of EEG features for the subjects, the cognitive load on each of the subjects is determined. For this, the cognitive load is classified under a predefined load class through a fuzzy classifier. In an implementation, the classification of the cognitive load of a subject is based on evaluating predefined fuzzy membership functions defined under the predefined load classes for the each predefined EEG feature for the subject. The predefined load classes are in conjunction with the predefined complexity levels of the predefined logical tasks. For example, the load classes may include low, medium, and high classes corresponding to low, medium, and high complexity levels, respectively. The fuzzy classifier classifies the cognitive load, for the subjects, under one of the predefined load classes with the help of predefined fuzzy membership functions of the selected EEG features.

[0019] In an implementation, the predefined membership functions under predefined load classes are determined for each common discriminative EEG feature identified for the test- subjects. The determined fuzzy membership functions for the test-subjects are used as a premise for classification of the cognitive load in the actual subjects.

[0020] The devices and methods of the present subject matter provide for measurement of cognitive load of subjects, while the subjects are executing logical tasks. Each logical task presented to the subjects is based on one or more propositional statements, which may be inferred as models of real-world tasks performed by individuals. The cognitive load thus determined upon presenting logical tasks based on propositional statements to the subjects is more reliable and realistic. Further, since the logical tasks based on propositional statements are presented to the subjects, the brain lobes responsible for logical thinking, for example, the frontal lobe part of brain, are stimulated, and hence the features extracted and subsequent analysis address the subjects cognitive abilities due to logical thinking.

[0021] Since, the methodology considers logical abilities of the subjects, the cognitive loads on the subjects are determined with substantially higher accuracy. For example, an individual at a work place is typically faced with challenges pertaining to logical thinking abilities, and therefore the cognitive load thus determined based on providing logical tasks are more realistic. Further, although the classification of cognitive load is associated with overlap of boundaries between two or more cognitive load levels or may vary with time due to dynamic logical abilities of the subject, the fuzzy classifier ensures distinctive classification of cognitive loads on the subjects under one of the predefined load classes.

[0022] The manner in which the devices and methods shall be implemented has been explained in details with respect to Fig. 1 and 2. Methods can be implemented in devices that include, but are not limited to, desktop computers, hand- held devices, laptops or other portable computers, mobile phones, and the like, capable of processing the EEG signal and determining the cognitive load. Although the description herein is with reference to computing devices, the methods and devices may be implemented in other devices and systems as well, albeit with a few variations, as will be understood by a person skilled in the art. While aspects of described devices and methods can be implemented in any number of different computing devices, transmission environments, and/or configurations, the implementations are described in the context of the following device(s).

[0023] Fig. 1 illustrates a system environment 100 implementing a cognitive load determination device 102, in accordance with an embodiment of the present subject matter. For the purpose of description and simplicity, the cognitive load determination device 102 is hereinafter referred to as a device 102. The device 102 can be implemented as a computing device, such as a laptop computer, a desktop computer, a notebook, a workstation, a mainframe computer, and the like. The device 102 is enabled to determine cognitive load on a subject 104, through Electroencephalography (EEG), while the subject 104 is executing logical tasks.

[0024] As shown in Fig. 1, an EEG measurement device 106 is coupled to the subject 104 to capture EEG signals for the subject 104 while the subject is performing or executing the logical tasks. The EEG measurement device 106 may be coupled to the subject 104 to capture the EEG signals from one or more brain lobes of the subject 104, which are responsible for logical thinking. Further, the device 102 is communicatively couple to the EEG measurement device 106 to obtain the EEG signals associated with the subject 104. In an implementation, the EEG signals may be obtained in analog or digital form, via a wired communication link, such as a data cable, or via a wireless communication link, such as Bluetooth™, IR, and WiFi. The device 102 and the EEG measurement device 106 are, respectively, equipped with compatible I/O interfaces for such communication.

[0025] In an implementation, the device 102 includes processor(s) 108. The processor(s) 108 may be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, state machines, logic circuitries, and/or any devices that manipulate signals based on operational instructions. Among other capabilities, the processor(s) is configured to fetch and execute computer-readable instructions stored in the memory.

[0026] The device 102 also includes interface(s) 110. The interface(s) 110 may include a variety of machine readable instruction-based and hardware-based interfaces that allow the device 102 to interact with other devices, including servers, data sources and external repositories. Further, the interface(s) 110 may enable the device 102 to communicate with other communication devices, such as network entities, over a communication network.

[0027] Further, the device 102 includes a memory 112. The memory 112 may be coupled to the processor(s) 108. The memory 112 can 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.

[0028] Further, the device 102 includes module(s) 114 and data 116. The module(s) 114 and the data 116 may be coupled to the processor(s) 108. The modules 114, amongst other things, include routines, programs, objects, components, data structures, etc., which perform particular tasks or implement particular abstract data types. The modules 114 may also be implemented as, signal processor(s), state machine(s), logic circuitries, and/or any other device or component that manipulate signals based on operational instructions. The data 116 serves, amongst other things, as a repository for storing data that may be fetched, processed, received, or generated by the module(s) 114. Although the data 116 is shown internal to the device 102, it may be understood that the data 116 can reside in an external repository (not shown in the figure), which may be coupled to the device 102. The device 102 may communicate with the external repository through the interface(s) 110.

[0029] Further, the module(s) 114 can be implemented in hardware, instructions executed by a processing unit, or by a combination thereof. The processing unit can comprise a computer, a processor, a state machine, a logic array or any other suitable devices capable of processing instructions. The processing unit can be a general-purpose processor which executes instructions to cause the general- purpose processor to perform the required tasks or, the processing unit can be dedicated to perform the required functions. In another aspect of the present subject matter, the module(s) 114 may be machine-readable instructions (software) which, when executed by a processor/processing unit, perform any of the described functionalities. The machine-readable instructions may be stored on an electronic memory device, hard disk, optical disk or other machine-readable storage medium or non-transitory medium. In one implementation, the machine-readable instructions can be also be downloaded to the storage medium via a network connection.

[0030] In an implementation, the module(s) 1 14 include a stimulus module

118, an EEG acquisition module 120, a signal processing module 122, a load classifier 124, and other module(s) 126. The other module(s) 126 may include programs or coded instructions that supplement applications or functions performed by the device 102. In said implementation, the data 1 16 includes test data 128, load analysis data 130, and other data 132. The other data 132 amongst other things, may serve as a repository for storing data that is processed, received, or generated as a result of the execution of one or more modules in the module(s) 1 14.

[0031] The description hereinafter describes the determination of cognitive load on one subject 104 based on provisioning of one logical task to the subject 104. Although the description is for one subject 104 and one logical task; cognitive loads on multiple subjects when provided with multiple logical tasks can be determined in a similar manner.

[0032] In an implementation, for the purpose of determination of cognitive load on the subject 104, the stimulus module 1 18 provides a logical task associated with a predefined complexity to the subject 104. The logical task involves or is based on at least one propositional statement. The complexity level of the logical task based on the number of propositional statements involved in the logical task. For example, the logical task maybe of a low, a medium or a high complexity depending on whether the logical task involves a single, a paired, a ternary or higher propositional statement construct. In an implementation, the logical task may be displayed by the stimulus module 118 on the computing device implementing the device 102, and the subject 104 then may view the logical task through a user-interface in the computing device and perform the logical task. [0033] The logical task is provided to the subject 104 from predefined logical tasks of predefined complexity levels. In an implementation, the complexity levels of the predefined logical tasks may be validated by an independent set of individuals, for example, graduate engineers. For this, a feedback on the complexity of the predefined logical tasks may be obtained from the independent set of individuals. Based on the feedback, the complexity levels of the predefined logical tasks may be either validated or tuned, to set the predefined complexity levels. The validation of complexity levels of logical tasks, before providing them to the subject 104, facilitates substantially accurate determination of cognitive load on the subject 104.

[0034] Table 1 illustrates examples of a few logical tasks based on one or more propositional statements. The complexity level associated with each of the logical tasks is also mentioned in Table 1. It may be inferred from Table 1 that the predefined complexity of logical task 1, 2 and 3 is low, medium, and high, respectively. For logical task 1, the answer can be found by executing a single chain rule based If-then propositional statement. Therefore, the complexity level associated with the logical task 1 is low. Similarly, for the logical tasks 2 and 3, two and three propositional statements are to be executed in the logical tasks, in order to arrive at the right answer. Therefore, the complexity level associated with the logical tasks 2 and 3 are medium and high, respectively.

Table 1

[0035] Further, the EEG acquisition module 120 obtains EEG signals from the EEG measurement device 106 coupled to one or more brain lobes of the subject 104, while the subject 104 is executing of the logical task. In an implementation, the EEG signals may be captured from frontal brain lobe, through F7, FC5, F3, and AF3 channels of the EEG measurement device 106 coupled to the subject 104. In an implementation, the EEG signals may be captured from brain lobes, such as parietal lobe and temporal lobe through T7, P7, 01, 02, P8, T8, FC6, F4, F8, AF4, F7, FC5, F3, and AF3 channels of the EEG measurement device 106 coupled to the subject 104. The EEG signals thus captured by the EEG measurement device 106 may be obtained by the EEG acquisition module 120 over a communication link between the device 102 and the EEG measurement device 106. [0036] Further, based on the obtained EEG signals for the subject 104, the signal processing module 122 evaluates a predefined set of EEG features from the EEG signals. The predefined set of EEG features include, but not restricted to, at least one of a log variance, a Total Band Power (TBR), a Log Band Power Ratio (LBPR), and Hjorth parameters representing activity, mobility and complexity. The details of the EEG features are described below.

[0037] Let us consider an EEG signal in time-domain as a sequence of time- sampled signal denoted by x(n), where n denotes the time sample. The frequency- domain representation of x(n) is given by X(f), where f denotes the signal frequency X(f) . The total band power (TBP) is defined as:

TBP = y ² X²(fi) (1) where, fl and f2 are the lower and upper cut-off frequencies for a band of the EEG signal. For delta (δ) band (DB), f_t = OHz and f₂ = 4Hz. For theta (Θ) band (TB), f\ = 4Hz and f₂ = 8Hz. For alpha (a) band (AB), fi = 8Hz and f₂ = 13Hz. For beta (β) band (BB), fi = 14Hz and f₂ = 30Hz.

[0038] The Log band power ratio (LBPR) is defined as:

^■LBPR i/j) = log TBPi)/ log TBPj ) (2) where i, j G {DB, ΤΒ,ΑΒ,ΒΒ} and i≠ j and TBPj denotes TBP for the i^th band.

[0039] Further, the variance is defined as the variation of the time-domain sampled EEG signal x(ri) from the average value x_av of the time-domain sampled EEG signal x(n). The variance is defined as:

Variance = [(xl - x_av)² + (x2 - x_av)²+... +(xn - x_av)²]/ⁿ _> (3) where, xl, x2, ... xn are the values in the time-domain sampled EEG signal x(n). The log variance is the logarithmic value of variance defined by equation (3). [0040] Further, the Hjorth parameters are defined as:

where x'defines the derivative of time-domain sampled EEG signal x(ri). [0041] In an implementation, the predefined set of EEG features, which are to be evaluated from the EEG signals for the subject 104, are predetermined in the device 102. The description below provides the procedure for determining the EEG features that can be considered as the predefined set of EEG features.

[0042] In an example, for determining the predefined set of EEG features, the stimulus module 118 provides the predefined set of logical tasks to multiple test- subjects, and the EEG acquisition module 120 obtains EEG signals from multiple channels of EEG measurement device coupled to one or more brain lobes of the multiple test-subjects, when the multiple test-subjects are executing the logical tasks. Based on the EEG signals for the test-subjects, the signal processing module 122 evaluates EEG features including, but not limited to, log variance, Total Band Power (TBR), Log Band Power Ratio (LBPR) and Hjorth parameters representing activity, mobility, and complexity. The signal processing module 122 then identifies common discriminative EEG features based on the values of the EEG features evaluated from the EEG signals for the test-subjects. [0043] The value of an EEG feature evaluated from an EEG signal for a test- subject performing a logical task depends on the complexity level of the logical task. The common discriminative EEG feature may be identified based on a correlation between the predefined complexity levels of the predefined set of logical tasks with the obtained EEG signals for all the multiple-test subjects. For this, the signal processing module 122 plots the values of each evaluated EEG feature associated with each channel of the EEG measurement device coupled to the test-subjects on a 1 -dimension plot. The EEG feature is identified as a common discriminative EEG feature if the values on the plot are distributed into classes, with points in each class being associated with one of the predefined complexity levels of the logical tasks. In an example, with the predefined set of logical tasks of low, medium, and high complexity levels, if the values of an EEG features are distributed into three separate classes, with the bottom class having values of EEG features corresponding to low complexity level tasks, the middle class having values of EEG features corresponding to medium complexity level tasks, and the top class having values of EEG features corresponding to high complexity level tasks, then that EEG feature is identified as a common discriminative feature.

[0044] In an example, one or more of the EEG features evaluated from the

EEG signals for the test-subjects may be identified as the common discriminative features. Table 2 illustrates examples of common discriminative EEG features identified from the evaluated EEG features based on the EEG signals associated with the test-subjects when provided with the predefined set of logical tasks. The EEG signals are obtained from F7, FC5, F3, and AF3 channels of the EEG measurement device worn by the test-subjects. In Table 2 the EEG features identified as common discriminative features are marked as Ύ', and the rest of the EEG features are marked as Ή'. As shown, LBPR (θ/β) in F7 and FC5 channels, and LBPR (δ/α) in F7 and AF3 channels are identified as the common discriminative EEG features. The identified common discriminative EEG features are set and recognized as the predefined set of EEG features which are then used for the determining the cognitive load on the subject 104.

Table 2

[0045] Further, based on the evaluation of the predefined set of EEG features for the subject 104, the signal processing module 122 evaluates predefined fuzzy membership functions for each of the evaluated EEG features. The predefined fuzzy membership functions are defined under predefined load classes for classification of cognitive load, where the predefined load classes are in conjunction with predefined complexity levels of the logical tasks. For example, if the predefined logical tasks are of low, medium and high complexity, then the load classes includes low, medium and high load class for classifying the cognitive load. The evaluated values of the predefined fuzzy membership functions are stored in the load analysis data 130.

[0046] In an implementation, the predefined fuzzy membership functions, which are to be evaluated based on the evaluated EEG features for the subject 104, are predetermined in the device 102. The fuzzy membership functions are determined under different load classes, based on the common discriminative EEG features identified for the multiple test-subjects. It may be noted that the fuzzy membership functions determined based on the common discriminative EEG features for the test- subjects are thus used as a premise for the classification of cognitive load in the subject 104. The description below describes the procedure for determining the fuzzy membership functions that can be considered as the predefined fuzzy membership functions. [0047] Let us consider that the common discriminative EEG features, marked with Ύ' in Table 2, are identified from the EEG signals for the test-subjects. For the purposes of description herein, let the common discriminative EEG features LBPR(9/p) of F7 channel, LBPR(6/p) of FC5 channel, LBPR(p/5) of F7 channel, and LBPR(p/6) of AF3 channel be denoted by fl, f2, f3, and f4, respectively. As mentioned earlier, for each of the common discriminative EEG features fl to f4, the values of the associated EEG feature for the corresponding channel are distributed into class based on the complexity levels of the logical tasks provided to the test- subjects. Each class has defined boundary coordinates. For determining the fuzzy membership functions under different load classes, the signal-processing module 122 determines the lower bound and the upper bound of the boundary coordinates for each class of each common discriminative EEG feature. For the complexity levels of logical tasks being low, medium, and high, the load classes include a low load class, a medium load class, and a high load class. Table 3 illustrates the lower bounds and upper bounds of the class for the common discriminative EEG feature fl , f2, f3 and f4. Here, fl iowmin represents the lower bound of the bottom class on the 1 -dimension plot for the common discriminative EEG feature fl. Since the fl iowmin is for the bottom class that represents EEG feature values corresponding to low complexity level logical tasks, the fl iowmin is determined as a minimum value for low load class of cognitive load. Similarly, fli_0Wmax represents the upper bound of the bottom class that represents EEG feature values corresponding to low complexity level logical tasks on the 1 -dimension plot for the common discriminative EEG feature fl, and is determined as a maximum value for low load class of cognitive load. Similarly, fl midmin and fl midmax represent the lower and the upper bound of the middle class, and are determined as minimum and maximum values for medium load class of cognitive load. Further, flhighmin and flhighmax represent the lower and the upper bound of the top class, and are determined as minimum and maximum values for high load class of cognitive load. The values in Table 3 corresponding to the other common discriminative EEG features f2, f3, and f4 are similarly defined. The data thus generated is stored in the test data 128.

Table 3

[0048] Now, let fi denote the i common discriminative EEG feature and μθ^ίϊ) denote the fuzzy membership function of fi under the load class Cj with Cj G {low, medium, high}. The fuzzy membership functions for low, medium and high load classes are defined as:

mcdium O = β-0/-/Ί)72σ2 (g)

where k, f , σ, k' are parameters of the fuzzy membership functions determined based on minimum and maximum values of load classes for the EEG features fl to f4 as described under Table 3.

[0049] In an example, the parameters k and f can be realized for μl_0W(fi)based on a boundary condition for the ^(ίΐ) that for fi = fiiowmin, μ^(ίΐ)= 1 , and for fi = fiiowmax, ΐο_¾(ίϊ) = 0.7. Solving for this boundary condition realizes f as equal to fiiowmin* and k as:

[0050] Further, in an example, the parameters f and σ for πιε_£ϋιιπι(ιΊ) can be realized as f equal to mean of the values of fi associated with the medium complexity level of logical tasks and σ equal to the variance of the values of fi associated with the medium complexity level of logical tasks.

[0051] Further, in an example, the parameters k' and f can be realized for

based on a boundary condition for the in a manner similar described for

[0052] For the example described herein, twelve different fuzzy membership functions are determined as: m_ow(fl), μΐο ν (£2),

Mmedium (f2), μηιβ(Ηυιη(ί3), μηιε<ϋιιηι(£4), high(fl), high(f2)_s μ^β.), Mhi_gh(f4). The determined fuzzy membership functions are set and recognized as the predefined fuzzy membership functions which are then used for the determining the cognitive load on the subject 104.

[0053] Based on the example described above, it may be noted that four common discriminative EEG features fl, f2, f3, and f4 are evaluated as the predefined set of EEG features from the EEG signals for the subject 104, and fuzzy membership functions are evaluated as the predefined fuzzy membership functions for each of the four evaluated EEG features, i.e., fl, f2, f3, and f4, and under each of the three load classes, i.e., low, medium, and high. In one implementation, the fuzzy membership functions may be evaluated by parsing the values of EEG features, evaluated from the EEG signals of the subject 104, onto the predefined fuzzy membership functions.

[0054] Further, based on the evaluated predefined fuzzy membership functions for the subject 104, the load classifier 124 classifies the cognitive load on the subject 104 under one of the predefined load classes. In an implementation, for the classification of cognitive load, the load classifier 124 evaluates fuzzy rules under the predefined load classes based on the values of the EEG features evaluated for the subject 104 and checks for the fuzzy rules which are satisfied. For classifying the cognitive load under the low, medium and high load classes, the fuzzy rules are defined as:

FRiow^'- [ (fl lowmin≤ f 1≤ fl lowmax) and (f2|₀wmin≤ β < f2i_0Wmax) and (f3l_0Wmin≤ f3

≤f3iowmax) and and (fni_owmin < fn < fni_owmax) ]

FRmedium- [ (flmidmin≤ fl ≤ fl midmax) nd (f2_mid_min≤f2≤ f2_midmax) nd (f3_midmin≤

O < Omidmax) and and (fn_midmin < fn < fn_midmax) ]

F high: [ (flhighmin≤ fl≤ f highmax) nd (Ghighmin≤ f2 f2highmax) and (f3highmin≤ f3 < f3highmax) and and (fn_hi_ghmin < fn < fn_highmax) ]

where FRi_0W, FR_medi_Um, and FRhigh denote the fuzzy rules under the low, the medium, and the high load class, fl, f2, f3, .... , fn denote the n number of predefined EEG features evaluated from the EEG signals for the subject 104. Further, fiiowmin and fiiowmax denote the lower and the upper bounds of the bottom class for the i^th EEG feature evaluated for the test-subjects against the low complexity level logical tasks. Similarly, fi_midmin and fimidmax denote the lower and the upper bounds of the middle class for the i^th EEG feature evaluated for the test-subjects against the medium complexity level logical tasks, and fihighmin and fihighmax denote the lower and the upper bounds of the top class for the i EEG feature evaluated for the test-subjects against the high complexity level logical tasks.

[0055] Further, based on determination of the fuzzy rules which are satisfied by the values of the predefined EEG features for the subject 104, the load classifier 124 computes a firing strength corresponding to each of the fuzzy rules which is satisfied. For classifying the cognitive load under the low, medium and high load classes using the fuzzy rules defined above, the firing strengths of the fuzzy rules FRiow, FR_medium, and FR_hj_gh are defined as:

¾_iow = min^iow(fl), μΐονν(β),... m_ow(fn) (1 1 ) ^FSFRmedium (12)

^FSPRhigh -

(13) where FS_FRI0W, FSpRmedium, and FSFRhigh denote the fuzzy strengths for the fuzzy rules for the low, the medium, and the high load class, and fl, f2, f3, .... , fn denote the n number of predefined EEG features evaluated from the EEG signals for the subject 104. Further,

denotes the predefined fuzzy membership function for the low load class evaluated for the i^th EEG feature for the subject 104. Similarly, μ_π,ε_ίϋ_ιιη_ι(Α) denotes the predefined fuzzy membership function for the medium load class evaluated for the i^th EEG feature for the subject 104, and

denotes the predefined fuzzy membership function for the high load class evaluated for the i^th EEG feature for the subject 104. Further, the function min() in equations (11), (12) and (13) implies that the firing strength takes a value that is minimum out of the values of the fuzzy membership function for that load class.

[0056] Further, after computing the firing strengths for the fuzzy rules which are satisfied for the subject 104, the load classifier 124 classifies the cognitive load on the subject 104 based on the values of the computed firing strengths. In an implementation, the load class for which the firing strength is maximum governs the class of cognitive load on the subject 104.

[0057] The description below describes the procedure for classification of cognitive load on the subject 104 through an example. Let us consider that four EEG features fl, £2, f3 and f4, as described above based on Table 2, are evaluated from the EEG signals for the subject 104, and twelve fuzzy membership functions are evaluated by on parsing the values of the four EEG features values into the predefined membership functions as described above under the three load classes (low, medium, high). In an example, the values of the fuzzy membership functions are normalized to scale them between 0 and 1. Table 4 illustrates example values of the fuzzy membership functions evaluated for the four EEG features fl to f4. Table 4

[0058] Now, based on the values of the four EEG features fl to f4, the fuzzy rules for the low, the medium, and the high load class, as described earlier, are evaluated. Based on the fuzzy rules FRi_0W, FR_medi_Um, and FRhigh, the firing strengths for each fuzzy rules is computed through equations (11), (12), (13) . It can be inferred from the values of fuzzy membership functions in Table 4 that the firing strength FSFRI_OW ⁼ 0.1 (the value corresponding to μι₀νν(ί4)). Similarly, the firing strength FSpRmedium = 0.5 (the value corresponding to μ_η6άίυιη(Ά)), and the firing strength FSFRhigh ⁼ 0.2 (the value corresponding to -highCf 1))- Based on the values of the firing strengths, the cognitive load on the subject 104 is classified "medium", as the firing strength for the medium load class possess the maximum value.

[0059] Although the description herein describes the determination of cognitive load on one subject through the device 102, when the subject is provided with a single logical task of a predefined complexity level; in an implementation, the device 102 can determine cognitive load on more than one subject, and, in an implementation, the device 102 can determine cognitive load of one or more subjects by providing them with more than one logical task.

[0060] In an implementation, multiple predefined logical tasks may be provided, one-by-one, to the subject 104, and the cognitive load on the subject 104 may be determined for each of the logical tasks performed by the subject 104. With this, each subsequent logical task being provided to the subject 104 can be modulated over the complexity level depending on the cognitive load determined for the prior logical task performed by the subject 104. For example, if the cognitive load determined for the subject 104 is high when the subject 104 is presented with and performing a predefined logical task of a medium complexity, then the subsequent logical task presented to the subject 104 may be of a low complexity.

[0061] Further, in an implementation, the device 102 may provide the predefined logical tasks on user devices of the subjects over a network, for determination of cognitive load across the subjects. The user devices may include, but not limited to, desktop computers, hand-held devices, laptops, tablet computers, mobile phones, PDAs, Smartphones, and the like that may be capable of receiving the EEG signals from the EEG measurement device worn by the respective subjects. In said implementation, each subject can perform the logical tasks on his user device, and the EEG signals from the EEG measurement device worn by the subject may be transmitted to the user device. The device 102 may then obtain the EEG signals for the subjects from each of the user devices over the network.

[0062] In an implementation, the network may be a wireless or a wired network, or a combination thereof. The network can be a collection of individual networks, interconnected with each other and functioning as a single large network (e.g., the internet or an intranet). Examples of such individual networks include, but are not limited to, Global System for Mobile Communication (GSM) network, Universal Mobile Telecommunications System (UMTS) network, Personal Communications Service (PCS) network, Time Division Multiple Access (TDMA) network, Code Division Multiple Access (CDMA) network, Next Generation Network (NGN), Public Switched Telephone Network (PSTN), and Integrated Services Digital Network (ISDN). Depending on the technology, the network includes various network entities, such as gateways, routers; however, such details have been omitted for ease of understanding. [0063] Fig. 2 illustrates a method 200 to determine cognitive load on a subject, according to an implementation of the present subject matter. The cognitive load is determined through Electroencephalography, while the subject is executing a logical task. The order in which the method 200 is described is not intended to be construed as a limitation, and any number of the described method blocks can be combined in any order to implement the method 200, or any alternative methods. Additionally, individual blocks may be deleted from the method 200 without departing from the spirit and scope of the subject matter described herein. Furthermore, the method 200 can be implemented in any suitable hardware.

[0064] The method 200 may be described in the general context of computer executable instructions. Generally, computer executable instructions can include routines, programs, objects, components, data structures, procedures, modules, functions, etc., that perform particular functions or implement particular abstract data types. The method 200 may also be practiced in a distributed computing environment where functions are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, computer executable instructions may be located in both local and remote computer storage media, including memory storage devices.

[0065] Further, although the method 200 may be implemented in any computing device; in an example described in Figure 2, the method 200 is explained in the context of the aforementioned device 102, for the ease of explanation.

[0066] Referring to Figure 2, at block 202, a subject 104 is provided with a logical task. The predefined logical task is based on at least one propositional statement, and logical task provided to the subject, is from predefined logical tasks of predefined complexity level. In one implementation, the subject may include a software programmer involved in logical analysis of program codes. Further, the complexity of the logical task is based on the number of propositional statements, incorporated in the logical task. In one implementation, the device 102 may present the logical task to the subject 104.

[0067] At block 204, the EEG signals from the at least one brain lobe of the subject 104 are obtained while the subject 104 is executing the logical task. The EEG signals may be obtained from frontal brain lobe, through F7, FC5, F3, and AF3 channels of the EEG measurement device 106 coupled to the subject 104. In another implementation, the EEG signals may be captured from brain lobes such as parietal lobe and temporal lobe and from T7, P7, 01, 02, P8, T8, FC6, F4, F8, AF4, F7, FC5, F3, and AF3 channels of the EEG measurement device 106. In an implementation, the EEG signals are obtained by the device 102, while the subject 104 is executing the logical task.

[0068] At block 206, a predefined set of EEG features from the obtained EEG signals are evaluated for the subject 104. The predefined set of EEG features may include, but not limited to, log variance, Total Band Power, Log Band Power Ratio and Hjorth parameters. The EEG features are evaluated by the device 102. In an implementation, the predefined set of EEG features is determined based on providing predefined logical tasks of predefined complexity level to multiple test-subjects, obtaining EEG signals for the multiple test-subjects, evaluating a plurality of EEG features from the EEG signals, and identifying common discriminative EEG feature as the predefined set of EEG features. The common discriminative EEG features can be identified in a manner as described earlier in the description.

[0069] Further, at block 208, predefined fuzzy membership functions defined under predefined load classes are be evaluated for each of the EEG features. The predefined load classes are in conjunction with predefined complexity levels of the logical tasks, provided to the subject 104. The fuzzy membership functions are evaluated by the device 102. In one implementation, the predefined fuzzy membership functions are determined based on the common discriminative features, identified for the test-subjects. The fuzzy membership functions can be determined in a manner as described earlier in the description.

[0070] At block 210, the cognitive load for the subject 104 is classified under one of the predefined load classes based on the predefined fuzzy membership functions. The cognitive load is classified by the device 102. In one implementation, the classification of the cognitive load includes, evaluating fuzzy rules under the predefined load classes based on the values of the EEG features evaluated for the subject 104, and computing firing strengths of the satisfied fuzzy rules under the predefined load classes. The cognitive load of the subject 104 is then classified based on the computed firing strength. The evaluation of fuzzy rules, the computation of firing strengths and the classification of the cognitive is done in a manner as described earlier in the description.

[0071] Although implementations for methods and devices for determination of cognitive load through EEG are described, it is to be understood that the present subject matter is not necessarily limited to the specific features or methods described. Rather, the specific features and methods are disclosed as implementations to determine cognitive load through EEG, on the subjects, while executing a logic-based task.

Claims

I/We claim:

1. A method for determining cognitive load on a subject, the method comprising: providing, by a computing device, a logical task to the subject, wherein the logical task is based on at least one propositional statement and is from predefined logical tasks of predefined complexity levels; obtaining Electroencephalography (EEG) signals from at least one brain lobe of the subject while the subject is executing the logical task; evaluating, by the computing device, a predefined set of EEG features from the EEG signals for the subject; evaluating, by the computing device, predefined fuzzy membership functions defined under predefined load classes for each of the EEG features, wherein the predefined load classes are in conjunction with the predefined complexity levels of the predefined logical tasks; and classifying, by the computing device, the cognitive load for the subject under one of the predefined load classes based on the predefined fuzzy membership functions to determine the cognitive load.

2. The method as claimed in claim 1 further comprising providing a subsequent logical task to the subject, wherein a complexity level of the subsequent logical task is based on the cognitive load determined for the subject.

3. The method as claimed in claim 1, wherein the EEG signals are obtained from F7, FC5, F3, and AF3 channels of an EEG measurement device coupled to the at least one brain lobe, and wherein the at least one brain lobe is a frontal lobe of the subject.

4. The method as claimed in claim 1, wherein the EEG features comprise at least one of a log variance, a Total Band Power, a Log Band Power Ratio, and Hjorth parameters.

5. The method as claimed in claim 1 further comprising determining the predefined set of EEG features, wherein the determining comprises:

providing predefined logical tasks, each with a predefined complexity level, to a plurality of test-subjects; acquiring EEG signals for the plurality of test-subjects, while the test- subjects are executing the predefined logical tasks; evaluating EEG features including a log variance, a Total Band Power, a Log Band Power Ratio and Hjorth parameters; and identifying common discriminative EEG features as the predefined set of EEG features, wherein a common discriminative EEG feature is identified based on distribution of an EEG feature for the test-subjects into class associated with the predefined complexity levels of the predefined logical tasks.

6. The method as claimed in claim 5 further comprising determining the predefined fuzzy membership functions for the common discriminative EEG features under each of the predefined load classes based on the distribution of the EEG features for the test-subjects.

7. The method as claimed in claim 1 , wherein the classifying the cognitive load is based on a firing strength computed under the predefined load classes through the predefined fuzzy membership functions in each of the predefined load classes.

8. A cognitive load determination device (102) for determining cognitive load on a subject (104), the cognitive load determination device (102) comprising: a processor (108);

a stimulus module (118) coupled to the processor (108), to provide a logical task to the subject (104), wherein the logical task is based on at least one propositional statement and is from predefined logical tasks of predefined complexity levels;

an EEG acquisition module (120) coupled to the processor (108) to obtain Electroencephalography (EEG) signals from a frontal brain lobe of the subject (104) while the subject (104) is executing the logical task;

a signal processing module (122) coupled to the processor (108), to evaluate a predefined set of EEG features from the EEG signals for the subject (104); and

a load classifier (124) coupled to the processor (108), to classify the cognitive load for the subject (104) based on the predefined set of EEG features.

9. The cognitive load determination device (102) as claimed in claim 8, wherein the signal processing module (122) evaluates predefined fuzzy membership functions defined under predefined load classes for each of the EEG features, wherein the predefined load classes are in conjunction with the predefined complexity levels of the predefined logical tasks, and wherein the load classifier (124) classifies the cognitive load under one of the predefined load classes based on the predefined fuzzy membership functions to determine the cognitive load.

10. The cognitive load determination device (102) as claimed in claim 8, wherein the stimulus module (118) further provides a subsequent logical task to the subject (104), wherein a complexity level of the subsequent logical task is based on the cognitive load determined in the subject (104).

11. The cognitive load determination device (102) as claimed in claim 8, wherein the plurality of EEG features include at least one of a log variance, a Total Band Power (TBR), a Log Band Power Ratio (LBPR), and Hjorth parameters.

12. The cognitive load determination device (102) as claimed in claim 8, wherein the signal processing module (122) further determines the predefined set of EEG features, wherein for determining the predefined set of EEG features, the stimulus module (118) provides the predefined logical tasks, each with a predefined complexity level, to a plurality of test-subjects; the EEG acquisition module (120) acquires EEG signals for the plurality of test-subjects while the test-subjects are executing the predefined logical tasks; and the signal processing module (122) evaluates EEG features including log variance, Total Band Power, Log Band Power Ratio and Hjorth parameters, and identifies common discriminative EEG features as the predefined set of EEG features, wherein a common discriminative EEG feature is identified based on distribution of an EEG feature for the test- subjects into class associated with the predefined complexity levels of the predefined logical tasks.

13. The cognitive load determination device (102) as claimed in claim 12, wherein the signal processing module (122) further determines the predefined fuzzy membership functions for the common discriminative EEG features under each of the predefined load classes based on the distribution of the EEG features for the test-subjects.

14. The cognitive load determination device (102) as claimed in claim 9, wherein the load classifier (124) classifies the cognitive load based on computing a firing strength under the predefined load classes through the predefined fuzzy membership functions in each of the predefined load classes.

15. A non-transitory computer readable medium having a set of computer readable instructions that, when executed, cause a computing device to:

provide a logical task to a subject, wherein the logical task is based on at least one propositional statement and is from predefined logical tasks of predefined complexity levels;

obtain Electroencephalography (EEG) signals from at least one brain lobe of the subject while the subject is executing the logical task;

evaluate a predefined set of EEG features from the EEG signals for the subject;

evaluate predefined fuzzy membership functions defined under predefined load classes for each of the EEG features, wherein the predefined load classes are in conjunction with the predefined complexity levels of the predefined logical tasks; and

classify the cognitive load for the subject under one of the predefined load classes based on the predefined fuzzy membership functions to determine the cognitive load.