AU2006264222B2 - Measuring cognitive load - Google Patents
Measuring cognitive load Download PDFInfo
- Publication number
- AU2006264222B2 AU2006264222B2 AU2006264222A AU2006264222A AU2006264222B2 AU 2006264222 B2 AU2006264222 B2 AU 2006264222B2 AU 2006264222 A AU2006264222 A AU 2006264222A AU 2006264222 A AU2006264222 A AU 2006264222A AU 2006264222 B2 AU2006264222 B2 AU 2006264222B2
- Authority
- AU
- Australia
- Prior art keywords
- person
- cognitive load
- measuring
- task
- patterns
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
Landscapes
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
This invention concerns a method for measuring cognitive load of a person in performing a task. In other aspects the invention can be expressed as a computer and as software that are used to perform the method. The computer (50) has an interface having a plural number of unimodal input (20, 30 to 40) and output devices, and a cognitive load analyser (50). The analyser comprises a receiver to receive input data signal streams (25, 35 and 45) from respective devices (20, 30 and 40). A classifier (56 to 59), (66 to 69) and (76 to 79) is also provided to identify predetermined “meta-interaction patterns” from the streams (25, 35 and 45), and to weight the identified predetermined “meta-interaction patterns” to produce respective weighted outputs. A combiner (80) to fuse the outputs to produce a measure indicating the person’s cognitive load.
Description
Title Measuring Cognitive Load Technical Field 5 This invention concerns a method for measuring cognitive load. In other aspects the invention can be expressed as a computer and as software that are used to perform the method. The concept of cognitive load has been used in a variety of fields that deal with the human mind interacting with some external stimulants. The definition of 10 cognitive load is slightly different in each field. For instance, in pedagogical literature cognitive load refers to the total amount of mental activity imposed on working memory at any instance in time; while in ergonomics literature it is described as the portion of operator information processing capacity, or resources that are required to meet cognitive task demands. Each field provides different methods to measure 15 cognitive load. In this specification the phrase "cognitive load" is defined as in the cognitive psychology literature, and its meaning is not a measurement of attention span, stress, engagement or other external elements to a task. Cognitive load can be defined as the mental effort required for a particular person to comprehend or learn some material, 20 or complete some task. Cognitive load is relative to both the user and the task being completed, at any single point in time. Background Art Conventional methods for measuring cognitive load, include: 25 subjective measures, such as self-rating scales; physiological techniques, such as pupil dilatation and heart rate; task or performance based measures, such as critical error rates; and behavioural measures, such as speech disfluencies, self-talk etc. There are a number of problems with these methods for measuring cognitive 30 load, including: some of the methods are intrusive and disrupt the normal flow of interaction; some of the methods are physically uncomfortable for the user; when conducted in real-time, some of the methods are labour-intensive; 35 the data quality is potentially unreliable outside laboratory conditions; and the data quality can be affected by outside factors, such as user's stress level.
WO 2007/000030 PCT/AU2006/000914 2 At this time there is no way to objectively quantify cognitive load that can be applied uniformly across fields as a standard, or to allow comparison between subjects. In fact, historically, the most consistent results for cognitive load assessments have been achieved through self-rating subjective measures; these allow 5 users to describe in fine detail, their perceived level of cognitive load induced by various types of task. A multimodal interface is an interface where the user is able to select one or more unimodal input or output devices to operate the interface. Use of each device represents a single mode of input or output. Most interfaces offer at least two input 10 modes, keyboard and mouse. Multimodal interfaces typically offer more options, including for instance voice recognition and other less common modalities. Disclosure of the Invention In a multimodal interaction scenario, a user experiencing a high cognitive load 15 will show involuntary behavioural anomalies that the inventors have come to understand to be symptoms (both physical and otherwise) of cognitive overload. These symptoms may be caused by high load or actual symptoms that are displayed when dealing with a high load. Additionally, the user may also change the manner in which they operate the multimodal interface to communicate. For instance, users 20 may: Continually move their mouse over a series of choices for an inordinately long time; Start shaking involuntarily; Make more non-critical mistakes in their interaction; 25 Change their mind more frequently than usual; or, Use particular kinds of multimodal constructions more often; or Rely more on their more dominant modality than others. Many other behavioural changes are also observable, but all the changes of interest concern the user's multimodal production or communicative act, rather than 30 the content of the communication. The inventors refer these as "meta-interaction features". A "meta-interaction pattern" is a determined pattern of "meta-interaction features." Recording these behavioural anomalies is akin to capturing prosodic information, or disfluencies from a speech signal - but extended to multimodal 35 interaction input channels. While they are an inherent part of the interaction, these anomalies are rarely deliberate or conscious reactions. In consequence the inventors propose a method for measuring a person's cognitive load, comprising the steps of: 3 Receiving one or more input data streams from an interface being used by a person while performing a task; Identifying predetermined "meta-interaction patterns" from the one or more streams; 5 Weighting the identified predetermined "meta-interaction patterns" to produce respective outputs; and Fusing the outputs to produce a measure indicating the person's cognitive load. The interface may be a multimodal interface. 10 The method may be unobtrusively performed to provide a real-time assessment of a user's cognitive load without attaching any devices to a user's body. Furthermore the measure may be objective. The received input data streams may involve one or more of the following types of information: 15 Visual, including eye movements, facial expressions, and gestures, such as deictic hand gestures, hand shapes, gesture trajectories, circling, underlining and various gestural marking methods. Visual data stream may be infrared red images of the head, body or hand. Speech, including pitch and fundamental frequency and its associated 20 higher order statistics, sound emissions (such as self-talk, mumbling and other verbal sounds), speech rate, and information throughput. Interaction events such as mouse hovering, mouse trajectory, menu and button selections, and typing and input method speed. Task characteristics such as task complexity or difficulty, current stage 25 or state of the task, error rate, completion rate, and latency to respond. User characteristics such as gender, age and education. Physiological data such as heart rate and perspiration. The method may further comprise the step of selecting input data streams, "meta-interaction patterns", method of performing the step of weighting and/or 30 method of performing the step of fusing that are specific to the person. Further, the method may further comprise the step of selecting input data streams, "meta interaction patterns", method of performing the step of weighting and/or method of performing the step of fusing that are specific to the task. For example, different classifiers will be derived for each type of input stream, and not all patterns may be 35 present in every user. Therefore the classifiers will change for any user/task instance. The step of identifying predetermined "meta-interaction patterns" from the streams may involve the steps of identifying predetermined features in each stream 3a and then identifying patterns among the features. The features that are chosen depend on the available input streams and whether the system deems there to be a significant WO 2007/000030 PCT/AU2006/000914 4 pattern found in that input stream, for that user. The type and number of input streams is therefore flexible. The features are then recognised and fused together to acquire a reliable measure of the cognitive load. The fusion method will also depend on the available 5 feature weights and may have several levels of incremental fusion, where small sets of features that are available are combined together first, before being combined with all available features. The patterns may be identified using one of more of the following models: Gaussian mixture models. 10 Hidden Markov models. Linear discriminant analysis. Decision trees. Artificial neural networks. Other statistical or non-statistical models. 15 The fusion step may be based on a probabilistic framework. Specifically it might be computing the probability of a set of features (X) indicating certain cognitive load level (0) and is given by: M P( I X) P(C. I X)y P(W I X,C,.)P( I W,X,Cm) m=1 B 20 where: P(CAM7X is the predicted accuracy (prior knowledge) of the m-th classifier given the feature X; P(9 X, C,) is the probability of having W for the given classifier; 25 P(0 [ OX, C,,) models the confusability (error pattern) of the classifier; M is the total number of classifiers. Once the output signals from the classifiers are combined, the final cognitive load level (0*) is determined by 30 0* = argmax P(O I X) The method may involve a calibration phase where a user is given a set of standard tasks to perform that vary in complexity, such as starting with simple tasks and gradually becoming more difficult. The output measure from each task may be considered and, depending on the proximity of this measure to an optimized pre-set 35 cognitive load target level for that task and user context, the next task is verified for appropriateness or changed.
5 The method may then involve a measurement phase where the user is asked to perform a task and the resulting output produced by the system is compared to the calibrated results to produce a measure of cognitive load. The weightings used in the measurement phase may be user specific, based on 5 the outcome of the calibration phase. Patterns that are more pronounced in reflecting the actual cognitive loads may be given a higher weighting. Machine learning may be applied to weight features more heavily for particular users, say, if they give a better indication of the user's cognitive load fluctuations. Dynamic learning may be applied to improve the individual classifications for 10 new users. Parameters of the individual classifiers and the fusion can be adjusted, for example through Bayesian Predictive Learning, to better reflect the characteristics of the new modality data acquired from users. The input data stream specific to a particular task may correspond to a 15 deliberate input by the person to the multimodal interface in the course of performing the task. In another aspect the invention is a multimodal computer to measure a person's cognitive load while performing a task, comprising a user interface having one or more unimodal input and output devices, and a cognitive load analyser comprising: 20 a receiver to receive one or more input data signal streams from the one or more unimodal input devices; and a classifier to identify predetermined "meta-interaction patterns" from the one or more streams, and to weight the identified predetermined "meta-interaction patterns" to produce respective weighted outputs; and, 25 a combiner to fuse the outputs to produce a measure indicating the person's cognitive load. The interface may adapt its behaviour to implement strategies that reduce cognitive load from use of the interface, and therefore increase task performance. Some options for behavioural adaptation include changes to: 30 The materials used in the interaction; The volume of information flow; The format of communication; and The range of tasks presented to the user during periods of both high and low cognitive load. 35 One or more of the following input devices may be used by the interface: Video cameras to collect data about the user's appearance, expression, position, movement and speed.
WO 2007/000030 PCT/AU2006/000914 6 Infrared cameras to record the temperature of the user or user pattern detection. Eye gaze trackers to track the user's glances, including looking at input devices, dwell times and glancing away. These things might be indicative of 5 attentiveness or disruptions. Accelerometers to detect the user's movement when it is too slight to be detected by the camera. Several accelerometers may be placed on the user's seat, on the floor, in the mouse or another handheld device where it is not noticeable for the user. 10 Input device trackers and piezoelectric transducers to track the movements and pressure applied on each input device. Sound capture transducers such as microphones and microphone arrays to capture speech and other sounds, such as foot tapping, produced by the user, as well as ambient noise, self talk, and other involuntary voice sounds. 15 Ultrasound to detect different kinds of spatial positioning of the user limbs and devices. The input devices used by the interface may be those input devices deliberately used by the person in performing the task. The interface may also involve a set of feature extractors to extract 20 predetermined features, for example time or frequency domain analysis, from the incoming data streams. In this case the classifiers identify predetermined "meta interaction patterns" from the features. The classifiers may identify the patterns using one of more of the following models: 25 Gaussian mixture models. Hidden Markov models. Linear discriminant analysis. Decision trees. Artificial neural networks. 30 Other statistical or non-statistical models. The classifiers may not be of the same model type, and a combination of model types within each feature stream and among different classifiers can be beneficial. Input data signals streams received by the receiver, the "meta-interaction patterns" and weights of the classifier and fusion of the combiner may be specific to 35 the person and/or task. In a further aspect the invention is software operable when installed on a computer to interface with a user to measure the user's cognitive load.
WO 2007/000030 PCT/AU2006/000914 7 Brief Description of the Drawings An example of the invention will now be described with reference to the accompanying drawings, in which: Fig. 1 is a block diagram of a multimodal interface of a computer; 5 Fig. 2 is a block diagram of a cognitive load analyser. Best Modes of the Invention The method of measuring the cognitive load typically takes place at a multimodal user interface of a computer. A task may be performed at the interface, or 10 alternatively, the interface may assist in human-human interaction. Referring now to Fig. 1, the multimodal interface of a computer 10 offers a plural number of input and output modalities. The input modalities include a camera 20 to produce visual data, a microphone 30 to produce audio data and conventional input devices such as keyboard and mouse 40 to produce data. Other, more exotic, 15 input modalities may be involved, such as speech recognition, eye-movement and gesture recognition. A more complete list of the modalities that can be used for measuring the cognitive load of a user is: Visual, including eye movements, facial expressions, head, body and 20 hand gestures, as well as infrared red images of the head, body or hands. Speech, including pitch and fundamental frequency and its associated higher order statistics, sound emissions (self-talk, mumbling) and speech rate. Interaction events such as mouse hovering over selections, toggling between selections and typing and input method speed. 25 Task characteristics such as task complexity or difficulty, current stage or state of the task and completion rate. User characteristics such as gender, age and education. Physiological data such as heart rate and perspiration. A more complete list of the sensor technologies that can be used to record the 30 modalities is: Video cameras to collect data about the user's appearance, expression, position, movement and speed. Infrared cameras to record the temperature of the user or user pattern detection. 35 Eye gaze trackers to track the user's glances, including looking at input devices, dwell times and glancing away. These things might be indicative of attentiveness or disruptions.
WO 2007/000030 PCT/AU2006/000914 8 Accelerometers to detect the user's movement when it is too slight to be detected by the camera. Several accelerometers may be placed on the user's seat, on the floor, in the mouse or another handheld device where it is not noticeable for the user. 5 Input device trackers and piezoelectric transducers to track the movements and pressure applied on each input device. Sound capture transducers such as microphones and microphone arrays to capture speech and other sounds, such as foot tapping, produced by the user, as well as ambient noise, self talk, and other involuntary voice sounds. 10 Ultrasound to detect different kinds of spatial positioning of the user limbs and devices. All sensors should be as inconspicuous as possible, so that they do not interfere with the user's interaction with the system. Each sensor will feed the captured data as a data stream, in this example three data streams are produced: a visual data stream 15 25, an audio data stream 35 and an input device data stream 45. The computer 10 also comprises a cognitive load analyser 50 which receives the data streams 25, 35 and 45, and this is shown in more detail in Fig. 2. The cognitive load analyser may be the processor of the computer 10 that is programmed using software to operate in the manner described below. 20 The cognitive load analyser 50 operates to receive input data streams 25, 35 and 45 at dedicated pre-processors (not shown) that operate to clean up the signals from the sensors 20, 30 and 40. The clean signals are then passed to feature extraction modules 55, 65 and 75 that operate to identify epochs of interest and extract pre-determined signal features 25 from the epochs of interest. Downstream of each feature extraction module, there are a number of classifiers; so downstream of the visual feature extraction module 55 there are classifiers 56, 57 - 59, and so on. These classifiers operate to identify predetermined patterns of features from each stream. For instance, co-occurring features in different 30 incoming signal streams can be 'bound' together to form a pattern. Various models can be used to build the classifiers. Some of the more feasible ones include Gaussian mixture models, hidden Markov models, linear discriminant analysis, decision trees, and artificial neural networks. In addition, the classifiers need not be of the same model type and a combination of model types within each 35 feature stream and among different classifiers can be beneficial. The classifier detects predetermined patterns for the task that the user is performing. That is, the patterns may be different for different tasks. The patterns are then each given a user specific weighting and the classifiers produce an output signal WO 2007/000030 PCT/AU2006/000914 9 related to the features they receive. Patterns that are more pronounced in reflecting the actual cognitive loads of the user are given a higher weighting. Once the features are weighted for that user, the output signals from the classifiers are combined together in fusion module 80 to acquire a measure of that user's cognitive load. 5 The fusion mechanism is based on a probabilistic framework that combines the outputs signals from various classifiers together, that is the probability of a set of features (X) indicating certain cognitive load level (0) and is given by: P(0\X)= P(Cm I X)yP(U X,Cm )P( IO,X,Cmn) m=1 9 10 where: P(CapX) is the predicted accuracy (prior knowledge) of the n-th classifier given the feature X; P(0 I X, C,) is the probability of having W for the given classifier; 15 P(O 10, X, Cm) models the confusability (error pattern) of the classifier; Mis the total number of classifiers. Once the output signals from the classifiers are combined, the final cognitive load level (0*) is determined by 20 0* = argmax P( I X) 0 In general there will be a calibration phase where a user is given a set of standard tasks to perform that vary in complexity, starting with simple tasks and gradually becoming more difficult. The output measure from each task is considered 25 and, depending on the proximity of this measure to an optimized pre-set cognitive load target level for that task, the next task is verified for appropriateness or changed. During a subsequent measurement phase the user is asked to perform a task and the resulting output produced by the system is compared to the calibrated results to produce a measure of cognitive load. 30 Although the invention has been described with reference to a particular example, it should be appreciated that many variations and modifications are possible. For instance, the selection of user modality characteristics could be changed to any number of permutations and combinations. The user modalities could be selected to suit the user and/or the task. Also, using Machine Learning or other techniques, 35 certain features could be weighted more heavily for particular users, say, if they give a better indication of the user's cognitive load fluctuations.
WO 2007/000030 PCT/AU2006/000914 10 In order to train the system uniformly, a set of control tasks representative of the full range of possible induced cognitive load is required. Constructing such a set is a difficult task given the number of variables, and each user's prior knowledge and contexts. The invention may address this issue by incorporating dynamic learning in 5 the measurement mode to improve the individual classifications for new users. Parameters of the individual classifiers and the fusion can be adjusted, for example through Bayesian Predictive Learning, to better reflect the characteristics of the new modality data acquired from users. In addition, the benefits can also be used for future new users: if certain new 10 tasks for one user are found to be inducing distinguished modality characteristics from the existing classifiers, then these tasks can be incorporated into the baseline training process to enhance the coverage of the training data. The collection and exchange of such data can be processed by a software module, possibly interconnected to other such modules by a data connection. 15 A side effect of measuring the cognitive load experienced by the user during each task could also be an indication of the cognitive load induced by different presentation strategies. The cognitive load measurement could be employed for uses other than context-aware user modelling, for example: 20 Applying cognitive load variables into online task assignment. Changing the presentation of the proceeding tasks in favour of preferred user modalities. Breaking down the proceeding tasks into smaller, ordered sub-tasks. Holding off or queuing less important or external tasks in preference of 25 more important tasks. Subtly altering the environment such that distractions are minimized. A further example of the invention will now be described with reference to Fig. 2. In this example there are two multimodal input streams, being visual data 25 and speech data 35. Speech data 35 is captured by a microphone and fed through a 30 speech recognizer. Gesture 25 is captured through video. Both of these inputs are connected and provided to a computer for further processing by the classifier and combiner as described below. The classifier and combiner are part of the processor of the computer, that is the processor is programmed by software to perform the functions of the classifier and combiner. 35 The task given to the user is a map-based task where the user is required to identify a number of places of interest on a map. The user is able to use both speech and gesture commands in isolation or in combination. The recognisers are able to WO 2007/000030 PCT/AU2006/000914 11 store and filter the signals for feature information of interest and are chronologically synchronised to allow comparison of signals across modality streams. Next, feature extraction is performed by the processor of the computer. The captured signals 25 and 35 are then sent through to the feature extraction modules 55 5 and 65. The features in the gesture input that may be captured are: Degree of shakiness in the gesture while pointing The size of the 'circles' used in selection (when circling an item) The extent to which gesture is used in the entire interaction (e.g. in 35% of constructions) 10 The features in the speech input that may be captured are: Delta-pitch (change in pitch at adjacent time sample) The frame energy in the speech signal The number of false-starts in the spoken command The number of hesitations in the spoken command 15 The features of combined multimodal use that may be captured are: The relative temporal overlap between speech and gesture input The quantitative temporal overlap between speech and gesture input Next, a series of classifiers are then able to identify pre-determined patterns in the signal and gauge how well this feature fits the pre-determined pattern. The data is 20 passed from the feature extraction modules 55 and 65 to the classifiers 56 to 59 and 66 to 69 respectively. The determined patterns that the classifiers are comparing the extracted features to are particular to the user and task and will have been extracted from a calibration phase or from a profile type that is appropriate for that user. 25 For example, when analysing the shakiness of the user's gesture input, the classifier may be just a threshold, where the shakiness is over X pixels, it becomes a point of interest. The classifier will be able to identify and mark the times when such events exceed the threshold. In addition, some classifiers may be combined requirements for two or more signals; e.g. the classifier can look for patterns of 30 exceeding the gesture shakiness threshold coinciding with a hesitation in the speech signal. The extent to which these modal signals fit the patterns determines the weighting of that feature. These results will be weighted at several levels, for example at the lowest level, if the signal is much higher than the threshold, then it is 35 given greater weighting. In addition, some signals will be weighted more strongly simply because they are more indicative of cognitive load fluctuations for that user. For example, for this user, when the shakiness exceeds the threshold more than 5 times in a minute, that is a strong indication of high cognitive load. For another user WO 2007/000030 PCT/AU2006/000914 12 this may be 10 times. The weighting can also be related to channel validity, for example, a weak signal that may be corrupt will automatically be given a lower weighting despite its significance to the user. Combined classifiers will produce again higher weightings than single modality classifiers. 5 The output from the classifiers is then passed to the fusion module 80. The fusion mechanism is based on a probabilistic framework that combines the outputs signals from various classifiers together. These can be fused at various levels of processing, and may have several levels of incremental fusion, where small sets of features that are available are combined together first, before being combined with all 10 available features. For example, all gestural based features will be fused together first, as well as all speech based features, then the output from these will be fused at a second level. Some features may not be used at all, if they do not contribute to the measurement of load, and have not reflected a sufficient change in to indicate a 15 fluctuation in load. The more modal features are used the higher the level of confidence the end result will have. The end construct may also include an indication of the modalities/modal features which provided more evidence of load fluctuations and this may be considered in choosing the next task or adjusting the user's profile to apply a higher 20 weighting to that modal feature. It is an advantage of this embodiment that input data signals are received from modal channels that are deliberately employed by the person completing the task. In this way the input data directly relates to the completion of the task, rather than non deliberate incidental actions such as head nods or other indicators of social activity. 25 It is a further advantage that particular input data streams are chosen specifically for the person performing the task. Further, the patterns and weightings can also be user specific. In this way an input data stream and be fused by the combiner in combination with other input data streams and/or in combinations with other input data streams. 30 It will be appreciated by persons skilled in the art that numerous variations and/or modifications may be made to the invention as shown in the specific embodiments without departing from the spirit or scope of the invention as broadly described. The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive. 35 Industrial Application Cognitive load measuring is important to many situations, especially high intensity work situation and long distance education.
WO 2007/000030 PCT/AU2006/000914 13 The scenario of high intensity work situations is typified by the magnitude of information required by the operators to make decisions that affect parts or the total environment under their control. Some examples include airplane take-off and landing, fire service, rail service, transportation management operators, ambulance 5 service and others. The workers need to make assessments in a matter of seconds and react appropriately, guided by the organisation's operational procedures. Incidents handled can be crisis situations where human life is in danger. In such a high intensity, stressful situation, the operators need to be able to interact seamlessly with the 10 technology available to them. Instead of hindering their access to information, the system should make it as easily accessible as possible, optimising the throughput of the information bandwidth between the user and the system; while allowing the user to remain in control of the interaction. If the system were able to reconfigure the interface to manage the flow of information to the user, depending on the user's 15 cognitive state and current situation, the workers would be able to handle their tasks more effectively. The real time assessment and management of the user's cognitive load as detected by the system would enable such technological environments to be designed. The learning cognitive load is a measure of the mental processing demand 20 involved in learning tasks. When this demand is too high or low, learning is ineffective. The development of a voice analysis-based learning difficulty diagnosis program, which measures learning cognitive load, would contribute immediately to the improvement of education for rural and dispersed learners, who participate in tele teaching, audio-conferencing, videoconferencing or who call help services, such as a 25 help desk. The outcomes of this project could supply the teacher, tutor or help desk operator clearer information about students' learning difficulties than currently available diagnostic methods. It could also be used to diagnose learning difficulties in conventional educational settings, and to assist in the development of instructional programs. Implementations of this invention could help provide both teachers and 30 students immediate feedback about their learning effectiveness.
Claims (21)
1. A method for measuring a person's cognitive load, comprising the steps of: receiving one or more input data streams from an interface being used by a 5 person while performing a task; identifying predetermined "meta-interaction patterns" from the one or more streams; weighting the identified predetermined "meta-interaction patterns" to produce respective outputs; and 10 fusing the outputs to produce a measure indicating the person's cognitive load.
2. A method for measuring a person's cognitive load according to claim 1, wherein the interface is a multimodal interface. 15
3. A method for measuring a person's cognitive load according to claim I or 2, wherein the received one or more input data streams involve one or more of the following types of information: visual data, including data that captures eye movements, facial expressions, head, body and hand gestures; 20 speech data, including pitch and fundamental frequency and its associated higher order statistics, sound emissions, speech rate; and information throughput; interaction events such as mouse hovering, mouse trajectory, menu and button selections and typing and input method speed; task characteristics such as task complexity or difficulty, current stage or state 25 of the task, error rate, completion rate, and latency to respond; and physiological data such as heart rate and perspiration.
4. A method for measuring a person's cognitive load according to any one of the preceding claims, wherein the step of identifying predetermined "meta-interaction 30 patterns" from the one or more streams comprises the steps of identifying predetermined "meta-interaction features" in each input data stream and then identifying patterns among the features.
5. A method for measuring a person's cognitive load according to claim 4, 35 wherein the patterns are identified using one of more of the following models: gaussian mixture models; hidden Markov models; linear discriminant analysis; 15 decision trees; and artificial neural networks.
6. A method for measuring a person's cognitive load according to any one of the 5 preceding claims, wherein the method further comprises the steps of any one or more of selecting one or more input data streams, "meta-interaction patterns", method of performing the step of weighting and method of performing the step of fusing that are specific to the person and/or specific to the task. 10
7. A method for measuring a person's cognitive load according to any one of the preceding claims, wherein the method further comprises a calibration phase where the person is given a set of standard tasks to perform that vary in complexity.
8. A method for measuring a person's cognitive load according to claim 7, 15 wherein the output measure from each task is considered and, depending on the proximity of this measure to an optimized pre-set cognitive load target level for that task and person context, the next task is verified for appropriateness or changed.
9. A method for measuring a person's cognitive load according to claim 8, 20 wherein the method further comprises a measurement phase where the person is asked to perform a task and the resulting output produced by the system is compared to the calibrated results to produce a measure of cognitive load.
10. A method for measuring a person's cognitive load according to any one of the 25 preceding claims, wherein the weightings used in the measurement phase is person and/or task specific.
11. A method for measuring a person's cognitive load according to any one of the preceding claims, wherein patterns that are more pronounced in reflecting actual 30 cognitive loads are given a heavier weighting.
12. A method for measuring a person's cognitive load according to any one of the preceding claims, wherein machine learning is applied to weight features more heavily for particular persons if they give a better indication of the person's cognitive 35 load fluctuations. 16
13. A method for measuring a person's cognitive load according to any one of the preceding claims, wherein dynamic learning is applied to improve the individual patterns for new persons. 5
14. A method for measuring a person's cognitive load according to any one of the preceding claims, wherein parameters of individual classifiers and the fusion can be adjusted to better reflect the characteristics of the new modality data acquired from the user. 10
15 A method for measuring a person's cognitive load according to any one of the preceding claims, wherein the one or more input streams are collected from the user in an unobtrusive manner.
16. A method for measuring a person's cognitive load according to claim 6, 15 wherein each selected input data stream specific to a particular task corresponds to a deliberate input by the person to the interface in the course of performing the task.
17. A computer to measure a person's cognitive load while performing a task comprising an interface having one or more unimodal input and output devices, and a 20 cognitive load analyser comprising: a receiver to receive one or more input data signal streams from the one or more unimodal input devices; and a classifier to identify predetermined "meta-interaction patterns" from the one or more streams, and to weight the identified predetermined "meta-interaction 25 patterns" to produce respective weighted outputs; and, a combiner to fuse the outputs to produce a measure indicating the person's cognitive load.
18. A method for measuring a person's cognitive load according to any one of the 30 preceding claims, wherein the method further comprises: receiving input of a person's characteristics, such as gender, age and education, wherein the step of identifying predetermined "meta-interactions patterns" is further based on the person's characteristics. 35
19. A method for measuring a person's cognitive load as substantially herein described with reference to the accompanying drawings. 17
20. A computer to measure a person's cognitive load as substantially herein described with reference to the accompanying drawings.
21. Software operable when installed on a computer that can interface with a person to perform the method of measuring the person's cognitive load as substantially herein described with reference to the accompanying drawings.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
AU2006264222A AU2006264222B2 (en) | 2005-06-29 | 2006-06-28 | Measuring cognitive load |
Applications Claiming Priority (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
AU2005903441A AU2005903441A0 (en) | 2005-06-29 | Measuring cognitive load | |
AU2005903441 | 2005-06-29 | ||
AU2006264222A AU2006264222B2 (en) | 2005-06-29 | 2006-06-28 | Measuring cognitive load |
PCT/AU2006/000914 WO2007000030A1 (en) | 2005-06-29 | 2006-06-28 | Measuring cognitive load |
Publications (3)
Publication Number | Publication Date |
---|---|
AU2006264222A8 AU2006264222A8 (en) | 2007-01-04 |
AU2006264222A1 AU2006264222A1 (en) | 2007-01-04 |
AU2006264222B2 true AU2006264222B2 (en) | 2013-05-02 |
Family
ID=38961335
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
AU2006264222A Ceased AU2006264222B2 (en) | 2005-06-29 | 2006-06-28 | Measuring cognitive load |
Country Status (1)
Country | Link |
---|---|
AU (1) | AU2006264222B2 (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5689619A (en) * | 1996-08-09 | 1997-11-18 | The United States Of America As Represented By The Secretary Of The Army | Eyetracker control of heads-up displays |
US6061610A (en) * | 1997-10-31 | 2000-05-09 | Nissan Technical Center North America, Inc. | Method and apparatus for determining workload of motor vehicle driver |
WO2002033529A2 (en) * | 2000-10-14 | 2002-04-25 | Motorola, Inc. | System and method for driver performance improvement |
US20040037236A1 (en) * | 2002-08-20 | 2004-02-26 | Motorola, Inc. | Method and apparatus for selecting a user interface |
US20040222892A1 (en) * | 2003-05-06 | 2004-11-11 | University Of Pittsburgh Of The Commonwealth System Of Higher Education | Apparatus and method for postural assessment while performing cognitive tasks |
-
2006
- 2006-06-28 AU AU2006264222A patent/AU2006264222B2/en not_active Ceased
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5689619A (en) * | 1996-08-09 | 1997-11-18 | The United States Of America As Represented By The Secretary Of The Army | Eyetracker control of heads-up displays |
US6061610A (en) * | 1997-10-31 | 2000-05-09 | Nissan Technical Center North America, Inc. | Method and apparatus for determining workload of motor vehicle driver |
WO2002033529A2 (en) * | 2000-10-14 | 2002-04-25 | Motorola, Inc. | System and method for driver performance improvement |
US20040037236A1 (en) * | 2002-08-20 | 2004-02-26 | Motorola, Inc. | Method and apparatus for selecting a user interface |
US20040222892A1 (en) * | 2003-05-06 | 2004-11-11 | University Of Pittsburgh Of The Commonwealth System Of Higher Education | Apparatus and method for postural assessment while performing cognitive tasks |
Also Published As
Publication number | Publication date |
---|---|
AU2006264222A8 (en) | 2007-01-04 |
AU2006264222A1 (en) | 2007-01-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CA2655189C (en) | Measuring cognitive load | |
Chen et al. | Multimodal behavior and interaction as indicators of cognitive load | |
US9737255B2 (en) | Measuring cognitive load | |
JP5233159B2 (en) | Group emotion recognition support system | |
US20190239791A1 (en) | System and method to evaluate and predict mental condition | |
CA3157835A1 (en) | Method and system for an interface to provide activity recommendations | |
US20130096397A1 (en) | Sensitivity evaluation system, sensitivity evaluation method, and program | |
CN108922629A (en) | The screening and its application of brain function corelation behaviour normal form index | |
JP2024012497A (en) | Communication methods and systems | |
CN108186030B (en) | Stimulation information providing device and cognitive index analysis method for potential value test | |
KR102552220B1 (en) | Contents providing method, system and computer program for performing adaptable diagnosis and treatment for mental health | |
US11779275B2 (en) | Multi-sensory, assistive wearable technology, and method of providing sensory relief using same | |
Yin et al. | Towards automatic cognitive load measurement from speech analysis | |
JP2007087255A (en) | Information processing system, information processing method and program | |
AU2006264222B2 (en) | Measuring cognitive load | |
CN114052736B (en) | System and method for evaluating cognitive function | |
WO2022102721A1 (en) | Depression-state-determining program | |
WO2022168180A1 (en) | Video session evaluation terminal, video session evaluation system, and video session evaluation program | |
WO2019227690A1 (en) | Screening of behavioral paradigm indicators and application thereof | |
Zhao | RETRACTED ARTICLE: Big Data Analytics integrated AAC Framework for English language teaching | |
WO2022249462A1 (en) | Video analysis system | |
WO2022209416A1 (en) | Information processing device, information processing system, and information processing method | |
WO2023032058A1 (en) | Video session evaluation terminal, video session evaluation system, and video session evaluation program | |
WO2022230068A1 (en) | Video analysis program | |
JP7138998B1 (en) | VIDEO SESSION EVALUATION TERMINAL, VIDEO SESSION EVALUATION SYSTEM AND VIDEO SESSION EVALUATION PROGRAM |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
TH | Corrigenda |
Free format text: IN VOL 22, NO 2, PAGE(S) 190 UNDER THE HEADING PCT APPLICATIONS THAT HAVE ENTERED THE NATIONAL PHASE - NAME INDEX UNDER THE NAME NATIONAL ICT AUSTRALIA LIMITED, APPLICATION NO. 2006264222, UNDER INID (54), CORRECT THE TITLE TO MEASURING COGNITIVE LOAD. |
|
FGA | Letters patent sealed or granted (standard patent) | ||
MK14 | Patent ceased section 143(a) (annual fees not paid) or expired |