WO2019183728A1

WO2019183728A1 - Alertness level measurement by measuring typing speed on devices

Info

Publication number: WO2019183728A1
Application number: PCT/CA2019/050379
Authority: WO
Inventors: Marc Therrien
Original assignee: Neuro Summum Inc.
Priority date: 2018-03-27
Filing date: 2019-03-27
Publication date: 2019-10-03

Abstract

There is described a method for measuring an alertness level of a user. The method comprises detecting user typing events on an electronic device and measuring a time interval between each consecutive one of the user typing events. All user typing events which correspond to a time interval outside an inclusion range are excluded, where the inclusion range characterizes time intervals related to alertness, either using predetermined thresholds and determining them after calibration. An assessment of the alertness level is made using time intervals between user typing events which were not excluded. Sleeping information which maximize the alertness level can then be identified.

Description

ALERTNESS LEVEL MEASUREMENT BY MEASURING

TYPING SPEED ON DEVICES

CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims benefit or priority of U.S. provisional patent application 62/648,432, filed March 27, 2018, which is hereby incorporated herein by reference in its entirety.

BACKGROUND

(a) Field

[0002] The subject matter disclosed generally relates to alertness level measurement. More specifically, it relates to a system to measure an alertness level using a mobile device.

(b) Related Prior Art

[0003] Alertness is related to most physical and mental capacities. Decreased alertness can be related to underperformance and to many risks, including accidents. Measuring alertness has however been very challenging and tools that exist to estimate alertness require equipment or dedicated time slots that prevent measuring alertness levels regularly throughout the day in varying environments.

[0004] The most sensitive measure of alertness is in a scientific setting and involves using an electroencephalogram to look at brain activity, but this procedure cannot be performed in everyday life.

[0005] The next most sensitive scientifically proven tool to measure alertness is the Psychomotor Vigilance Task (PVT) that consists of a 10-minute serial reaction time test. Shorter versions of 3 minutes have also been showed to be valid to measure alertness levels. We have corroborated these findings in a study using a 3-minute reaction time test, and we have found that even the first 2 minutes of the test are sufficient to measure alertness. [0006] There is a mobile application called“sleep-2-Peak” that measures alertness by having the user do a reaction time test, which involves tapping on a geometrical shape as fast as possible when it appears on the screen. The application was studied in“Validation of sleep-2-Peak: A smartphone application that can detect fatigue-related changes in reaction times during sleep deprivation” (Brunet, JF. , Dagenais, D., Therrien, M. et al. Behav Res (2017) 49: 1460. https://doi.org/10.3758/s13428-016-0802-5).

SUMMARY

[0007] According to an aspect of the invention, there is provided a method for measuring an alertness level of a user, the method comprising:

- detecting user typing events on an electronic device;

- measuring a time interval between each consecutive one of the user typing events;

- excluding all user typing events which correspond to a time interval outside an inclusion range, where the inclusion range characterizes time intervals related to alertness;

- producing an assessment of an alertness level using time intervals between user typing events which were not excluded.

[0008] According to an embodiment, detecting user texting events on an electronic device comprises detecting user typing events on a mobile electronic device.

[0009] According to an embodiment, user typing events additionally include user screen touches performed outside of a keyboard of the mobile electronic device (i.e. , the user screen touches performed outside of a keyboard of the mobile electronic device are used in addition to the user typing events in the steps of detecting, measuring a time interval, excluding, and producing an assessment). [0010] According to an embodiment, there is further provided the step of providing a monitoring application on the mobile electronic device, wherein detecting user typing events is performed by the monitoring application.

[0011] According to an embodiment, detecting user typing events is performed when the monitoring application is not open and is performed for user typing events occurring in an application other than the monitoring application.

[0012] According to an embodiment, detecting user typing events comprises having the monitoring application monitor only typing performed on a keyboard of the mobile electronic device.

[0013] According to an embodiment, there is further provided the step of installing the monitoring application on the mobile electronic device, said installing comprising installing a custom keyboard for use in the application other than the monitoring application, wherein detecting user typing events comprises detecting user typing events performed on the custom keyboard.

[0014] According to an embodiment, detecting user typing events on an electronic device comprises detecting user typing events on a keyboard peripheral.

[0015] According to an embodiment, there is further provided the step of determining, prior to excluding, the inclusion range based on a calibration sequence of typing events, comprising determining a lower threshold and an upper threshold for the time interval.

[0016] According to an embodiment, determining a lower threshold and an upper threshold comprises determining the lower threshold and the upper threshold relative to an average time interval in the calibration sequence of typing events, the average time interval being computed as one of: a mean time interval of all typing events in the calibration sequence of typing events, a mean of the time interval of the typing events in the calibration sequence of typing events excluding outliers, a median time interval of all typing events in the calibration sequence of typing events, and a median of the time interval of the typing events in the calibration sequence of typing events excluding outliers.

[0017] According to an embodiment, determining a lower threshold and an upper threshold relative to an average time interval comprises setting the lower threshold at a range between 20% and 70% under the average time interval and setting the upper threshold at a range between 20% and 70% above the average time interval.

[0018] According to an embodiment, determining a lower threshold and an upper threshold relative to an average time interval comprises setting the lower threshold at a range between 40% and 60% under the average time interval and setting the upper threshold at a range between 40% and 60% above the average time interval.

[0019] According to an embodiment, determining a lower threshold and an upper threshold comprises setting the lower threshold between 100 and 220ms and setting the upper threshold between 300 and 500ms.

[0020] According to an embodiment, there is further provided the step of recording, within the monitoring application, sleeping information of the user.

[0021] According to an embodiment, recording the sleeping information of the user is performed by the user entering the sleeping information in a graphical user interface of the monitoring application.

[0022] According to an embodiment, recording the sleeping information of the user is performed by the monitoring application on a mobile electronic device detecting events in the environment of the mobile electronic device indicative of the sleeping information.

[0023] According to an embodiment, producing the assessment of the alertness level comprises averaging the time intervals between user typing events which were not excluded over a period to produce the assessment of the alertness level which is an average of the time intervals, not excluded, for the period.

[0024] According to an embodiment, the period for producing the assessment of the alertness level is between 0.5 hour and 2 hours and wherein the assessment is performed repeatedly.

[0025] According to an embodiment, the period for producing the assessment of the alertness level is less than 10 seconds.

[0026] According to an embodiment, there is further provided the step of identifying sleeping information of the user which contribute to higher alertness level characterized by a smaller average of the time intervals for the period.

[0027] According to an embodiment, there is further provided the step of determining, for a given period on a future day, which sleeping information, applied to a sleeping period preceding said future day, would induce the higher alertness level for the given period in the future day, and outputting a result of said determining.

[0028] According to another aspect of the invention, there is provided an electronic device comprising a memory for storing instructions and a processor in communication with the memory for executing the instructions which cause the electronic device to perform the steps of:

- detecting user typing events occurring in the electronic device;

- producing an assessment of an alertness level using time intervals between user typing events which were not excluded. [0029] According to an embodiment, the electronic device is a mobile electronic device and the typing events are tapping events on a screen of the mobile electronic device.

[0030] According to an embodiment, there is further provided a monitoring application executable by the processor of the mobile electronic device, wherein detecting user typing events is performed by the monitoring application.

[0031] According to an embodiment, there is further provided an application other than the monitoring application executable by the processor of the mobile electronic device, wherein detecting user typing events is performed when the monitoring application is not open and is performed for user typing events occurring in the application other than the monitoring application.

[0032] According to an embodiment, there is further provided a custom keyboard installed on the mobile electronic device and executable by the processor of the mobile electronic device for use in the application other than the monitoring application, wherein detecting user typing events comprises detecting user typing events performed on the custom keyboard.

[0033] According to an embodiment, there is further provided a keyboard peripheral connected to the electronic device for detecting user typing events thereon.

BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Further features and advantages of the present disclosure will become apparent from the following detailed description, taken in combination with the appended drawings, in which:

[0035] Fig. 1 is a schematic diagram illustrating a user typing on a keyboard in an application while the typing events are recorded by a monitoring application, according to an embodiment; [0036] Fig. 2 is a graph illustrating recorded typing events and their respective timestamps, according to an embodiment;

[0037] Fig. 3 is a graph illustrating recorded typing events and the texting intercharacter time interval (TICT) between each pair of consecutive typing events, according to an embodiment;

[0038] Fig. 4 is a screenshot of a user interface of the monitoring application for calibration, according to an embodiment;

[0039] Fig. 5 is a screenshot of a user interface of the monitoring application for entering sleep-related data, according to an embodiment;

[0040] Fig. 6 is a graph illustrating hourly averages of reaction times (points accumulated over a plurality of days), according to an embodiment;

[0041] Fig. 7 is a graph illustrating daily averages of reaction times (dual points for each one of a plurality of days) depending on sleeping times, according to an embodiment;

[0042] Fig. 8 is a graph illustrating hourly averages of both reaction times and TICTs (points accumulated over a plurality of days), showing a correlation between them, according to an embodiment;

[0043] Fig. 9 is a graph illustrating daily averages of both reaction times and TICTs (dual points for each one of a plurality of days) depending on sleeping times, showing a correlation between them, according to an embodiment; and

[0044] Fig. 10 is a flowchart illustrating a method for measuring an alertness level, according to an embodiment.

[0045] It will be noted that throughout the appended drawings, like features are identified by like reference numerals.

DETAILED DESCRIPTION

[0046] The measures described above in reference with the prior art (such as Brunet et al.) require that the individual stops their activities and concentrates on doing the test. This is annoying for the individual, who may choose to reduce the number of times they take the test each day, resulting in a dataset of decreasing size over time as individuals get tired of stopping their normal activities to take the test. This ends up being inefficient and may result in a less performing measurement of alertness level if the result is that less measurements are collected overall. Moreover, even in the case where the individual keeps taking the test regularly, having to do so is intrinsically cumbersome and annoying.

[0047] Being able to measure the alertness level while individuals continue their regular activities would be a significant advantage to track alertness levels and variations throughout the day and longitudinally. This would ensure a larger dataset since a greater number of measurements will be collected each day, and would also be collected more regularly (i.e. , there would be less“holes” in data acquisition across daytime).

[0048] We have found that physical motor speed variations are highly correlated to variations in the alertness level.

[0049] More specifically, we have found that the simple movements which are executed the fastest, i.e., those that have become mostly automatic, are those that are best correlated to reaction time and alertness levels.

[0050] The invention described herein assumes that texting speed, i.e., the rate at which characters are typed on a mobile electronic device, can be a valid and useful measure of the alertness level. This is especially valid for carefully selected series of consecutive strokes, as described further below.

[0051] Considering that users rarely take time to do a reaction time test throughout the day but that most people text or write using the keyboard of their smartphone regularly almost every day, it is considered that using measurements of variations in texting speed, if proven to be correlated to variations in reaction time, could be a very useful and potentially widespread tool to inform individuals about their alertness levels, their best sleep schedule and their best activity periods during the day to reach their best functioning levels.

[0052] Considering that alertness measurement using psychomotor abilities is best measured by using the fastest and most simple tasks, like simple reaction time, it is considered that for measuring alertness by using texting speed, focus should be given to measuring the most simple and direct components of texting. Thus, measuring the tapping time interval between two consecutive characters in an already mentally planned word or phrase (i.e. , a sequence of typing events), excluding periods where the user is most probably thinking, searching for a special character or any other mental activity (i.e., using only an interrupted sequence of typing events), is considered the best approach.

[0053] Having found the usefulness of texting speed to the measurement of alertness level, there is a need to measure texting speed appropriately, more particularly, to detect all screen touches when the user is using his mobile device, and ensuring that an interrupted sequence of typing events is used for making measurements, thus requiring a way to identify interruptions in a sequence of typing events.

[0054] According to an embodiment, when the monitoring application is activated and monitoring texting events in the other application 120 (i.e., an application other than the monitoring application 110), an icon 115, preferably of a small footprint on the screen, appears on the screen (e.g., in a corner). This is to notify the user than even though they are not explicitly using the monitoring application 110, the monitoring application 110 is currently monitoring the user activity with the other application 120. When the icon is present, the monitoring application is detecting screen touches, preferably every occurrence (i.e., continuously detecting over a period / interval of time / duration). The exact number of all screen touches, with a precision in the order of milliseconds (e.g., precision of 1 ms), will thus be detected and recorded by the monitoring application. [0055] The monitoring application 110 should be connected to the operating system of the mobile electronic device 10 to perform these recurrent detections. Examples of operating systems at the present time include, for example, iOS and Android (both trademarks), although other OSes exist on the market.

[0056] The records of all touches can be stored on the memory of the mobile electronic device 10, either temporarily (in a cache or app memory) or permanently (on a hard drive memory). In either case, the recorded data can be backed-up (i.e. , copied) or transferred to a remote server (e.g., an application server, distributed servers in the cloud or any other equivalent). Doing so can allow for an eventual deletion of accumulated data on the smartphone where the storage for such a large amount of raw data is usually limited. Alternatively, the log of daily average or other meaningful features in the collected data can be kept (on the phone and/or on remote servers) while the raw data of individual touches, from which these features are extracted, can be deleted to keep storage requirements reasonable over time and circumvent cybersecurity issues or privacy issues that could be associated to the storing of data with very fine granularity.

[0057] According to an embodiment, the monitoring application 110 monitors each and every screen touch. According to another embodiment, which may be preferred for greater accuracy and reducing processing and storage requirements, the monitoring of touches is only performed when the keyboard portion 125 of the mobile electronic device 10 is open on the user interface of the operating system of the mobile electronic device 10, and in this case, only touches made on the screen on the keyboard portion 125 of the mobile electronic device 10 are recorded, i.e., touches made on other parts of the screen not belonging to the keyboard portion 125 are excluded from the monitoring and recording by the monitoring application 110.

[0058] In another embodiment, it may be determined that on-screen tapping events outside of the keyboard portion 125 will naturally be filtered (using an exclusion algorithm described further below) and not used for averaging the texting intercharacter time interval (TICT) over a period, making the distinction between on-keyboard or off-keyboard tapping events less relevant during the step of collection of screen events.

[0059] According to an embodiment, only screen touches are monitored, while the identity of the character selected by the touch on the keyboard portion 125 is not monitored. Doing so ensures security and confidentiality, because for most users, it is important that the content of what is texted is not detected, measured or recorded. Detecting only the timing of screen touches, i.e. , only the time stamp associated to a screen touch, without detecting where the screen is touched on the keyboard thus assures total confidentiality and security.

[0060] In another embodiment, all data, including the exact location on the screen where the touch occurred, optionally including the character selected, are recorded for greater precision in identifying that a touch was made during the course of a texting sequence. However, this exemplary embodiment may be hard to implement due to other considerations such as respect of privacy, especially in view of cybersecurity issues that could arise from storing such sensitive data. Moreover, the computation power needed to collect all these data may be too much for a mobile device and may distort the time measurement, making the data collection not a real-time data collection (i.e., there would be a time offset between actual time of tapping and collected timestamp of the event).

[0061] According to another embodiment, the monitoring application may include a keyboard application used as a replacement for the default keyboard in the other applications. This replacement keyboard implements the collection capabilities mentioned above while respecting privacy (i.e., only timestamps are collected.

[0062] According to an embodiment, data will normally be collected in all applications which are not the monitoring application 110, except in the calibration steps (discussed further below) and for text naturally written within the monitoring application (e.g., user comments possibly tagged to collected data). In an embodiment, permissions are required before performing monitoring in a given application other than the monitoring application itself. Therefore, the monitoring application, even when not seen or not used by the user, operates in background in order to monitor texting events occurring in other applications 120, such as a text application, an email application, a web browser, a social media application, etc., possibly when the other application 120 has a keyboard used by the user although it may not be required depending on the OS environment.

[0063] According to an embodiment, the monitoring application 110 is in communication with the operating system of the mobile device to monitor typing events occurring in the native keyboard of the mobile device or of its operating system. This may include monitoring events in a keyboard that is not the native keyboard in the operating system, but is used in replacement thereof, in particular third-party keyboards which can be customized.

[0064] According to another embodiment, the monitoring application 110 can include a replacement keyboard which is installed on the operating system of the mobile device and is set to replace the native keyboard. This equivalent of a “third-party keyboard” but belonging to the monitoring application 110 can be set by the user as their preferred/installed keyboard on the mobile device. This may facilitate recording of keyboard typing events because the keyboard used in the other applications 120 now use the keyboard which belongs to the monitoring application 110.

[0065] Fig. 2 is a graph illustrating a record of all texting events recorded by the monitoring application 110. All texting events recorded are discrete and instantaneous events and have a timestamp associated thereto, as the monitoring application 110 queries the clock of the mobile electronic device for each detected touch event and thus records the timestamp for each detected texting event. [0066] Fig. 3 is a graph illustrating the texting intercharacter time (now referred to as the“TICT”), which is the time period or time interval between each texting or typing event, i.e. , the time interval from the last recorded texting event to the currently detected texting event. The time of day, or clock time, is queried at each detected texting event to attribute a timestamp for each event (as shown in Fig. 2); the TICT can be computed thereafter by subtracting the time between two consecutive events (as shown in Fig. 3). As described in further detail below, a selection of relevant events needs to be performed, where only the intervals between texting events related to alertness (i.e., less influenced by other mental activity and not involving distraction) are kept. A TICT outside of a certain range (i.e., outside an inclusion range) will be indicative of hesitations, a pause or other mental activity (distraction or other tasks where the out-of-range TICT is not related to alertness).

[0067] In an embodiment, only the time interval between texting events is actually recorded by the monitoring application, while the timestamp (i.e., time of day, or clock time) is only recorded a few times a day (e.g., once every hour or half hour).

[0068] In order to extract meaningful features from the collected data regarding the measurement of the alertness level, only relevant TICTs should be kept. In order to measure only the time intervals between characters that are considered to have been tapped consecutively without any technical or mental interference (i.e., hesitations, pauses, finding words, thinking about the sentence, etc.), it is considered that all “continuous” texting will be represented by intercharacter time intervals clustered in a short time interval group, and only these clusters will be averaged (i.e., arithmetic mean) for alertness assessment.

[0069] Even though typing is a sequence of discrete events, the“continuity” in the sequence refers to the absence of interruption such as hesitations or pauses, which can be identified and serve as filters for clustering events only in groups, each corresponding to an uninterrupted sequence. [0070] It is considered that all the other TICTs that are out of this cluster represent intercharacter time intervals where the user was engaged in another mental activity (not simply typing a series of consecutive characters), including thinking about what to write, hesitations about spelling, looking for a special character or emoji, text correction, etc. Measuring texting speed to measure alertness thus measures average intercharacter time using only this cluster (or plurality of concatenated time series clusters) of shortest intercharacter intervals, excluding the intercharacter time interval associated to interruptions.

[0071] There is thus performed a clustering of data, as shown in Fig. 3. The method should be able to collect all data and to categorize the collected data as being either: strokes belonging to the same uninterrupted sequence or series in which no other mental activity is performed, or strokes that mark a hesitation or other mental activity and which does not belong to the neighboring sequences of strokes (the plural referring to the sequences before and after the interruption).

[0072] For example, in an empirical setting in which typing events are collected for an individual, characters are written on a touchscreen of a smartphone with intercharacter time intervals ranging between 220ms and 310ms. In this example, no intercharacter time intervals were present between 310 and 540ms. The balance of intercharacter times are irregularly dispersed without a specific pattern between 540ms and 3245ms, for example for hesitations or pauses. (These numbers are only exemplary and vary depending on the individual.) These are shown in Fig. 3.

[0073] Therefore, in order to make hourly averages of intercharacter time intervals, only the intercharacter time intervals belonging to a range defined by a low and a high threshold are kept, as shown in Fig. 3, this range of values between the thresholds being the cluster that is kept. A plurality of clusters within a given period (in the order of a few minutes) can then be concatenated. [0074] However, interindividual texting speed differences can be significant, i.e. , different people have different texting speeds, which can be significantly different. Therefore, according to an embodiment, an initial calibration of average texting speed is performed for a new user (or an existing user using a new device). This initial calibration should be performed to identify a range of TICTs that will be considered afterwards to measure the relevant text typing speed, representative of the alertness level, i.e., the time intervals between strokes that will be included to or excluded from the selection of time intervals that are considered as indicative of the alertness level.

[0075] As shown in Fig. 4, for the calibration, the user is asked, within the user interface of the monitoring application 110, to text a few lines of words, some lines requiring the user to text a given series of words, and other lines being open answers to simple questions, for example. From these few lines of texting, the cluster of fastest intercharacter time intervals is determ ined and an average texting speed for that individual is defined. Instead of selecting the lower threshold and the upper threshold as 220ms and 310ms, respectively, one may define the threshold as the range set between 50% faster to 50% slower of the average time intervals for the series of characters typed during calibration, i.e., for the relevant series of characters clustered in the shortest intercharacter time intervals as presented as the texting sequences in Fig. 3. These would be the calibrated thresholds for the range of selected time intervals to be considered for the averaging and considered to be relevant for assessing a level of alertness.

[0076] According to an exemplary embodiment, from this average of intercharacter time intervals, a range set between 50% faster (lower TICT threshold) and 50% slower (upper TICT threshold) is delimitated as the range of intercharacter time intervals that will be retained in future texting to measure texting speed average for a time period. This range is used to select intercharacter time intervals that are relevant to the measurement of alertness levels and for which hourly and daily averages will then be computed and recorded. The monitoring application 110 can also calculate the evolution of the average TICT and reset the calibration average and thus the delimitations (avg. TICT ±50%) of the retained texting intercharacter times.

[0077] By using specific limited time windows throughout the day, for example all screen touches between 10:30am and 1 1 :30am, averaging the relevant TICTs during that time window gives the average TICT for that period. This is shown in Fig. 8. This graph shows the relationship between time of the day that the texting speed is being measured and the results of the average texting speed for every one-hour window (ex: average texting speed at 1 1 :00 am for all texting done between 10:30 and 1 1 :30). The choice of time intervals for averaging is discretionary and can be varied if needed, although one-hour windows (or half- hour or two-hour windows) are easy to understand and to apply in real life situations. This allows the user to become aware of the variations in alertness throughout the day and help the individual plan his activities, including to assure peak performance when needed.

[0078] Fig. 5 illustrates an embodiment of a user interface of the monitoring application 110 in which the user can enter their sleeping information such as their previous bedtime and out-of-bed hours, i.e. , enter a timestamp for going to bed time and for waking up time. According to a more specific embodiment, the user may further specify that other events occurred, especially those that affect alertness or sleeping, such as alcohol intake, sleeping pills, or coffee, as shown, and tag a timestamp to these events, as additional sleeping information which complement the sleeping times.

[0079] Bedtimes can also be estimated using user activity on the mobile electronic device 10 as a proxy for bedtime hours. The monitoring application 110 may then prompt the user to validate that these estimated bedtimes based on mobile electronic device activity are exact, or it may simply provide the user with an interface to modify these values. Alternatively, the bedtimes can be imported from another application or device dedicated to this task, e.g., from a smartwatch or monitoring watch (e.g., a Fitbit™) which can feed their data on sleep quality and sleeping hours to the monitoring application. Alternatively, the monitoring application can query the accelerometer of the mobile electronic device 10 to determine that the mobile electronic device 10 is being handled, this handling being indicative of awake time and thus being a proxy for bedtime hours.

[0080] The results of the data collection are then correlated to the reported sleep schedule of the preceding night or nights. Over time, the user can thus see on a graph (Fig. 7) which bedtimes and out-of-bed times are associated to their best reaction times, or equivalently their lowest TICT, during the day. Both metrics (traditional reaction time measurements and TICTs) are highly correlated, as shown in Fig. 9, although as discussed herein, measuring TICT presents several advantages in term of data quantity and time coverage. Another graph (Fig. 6), discussed above, allows the user to see at what time of the day he has his best reaction times and at what times his reaction times sub-optimal. Again, this metric is highly correlated to the TICT, as shown in Fig. 8 where both metrics are shown, and measuring TICT presents several advantages compared to traditional reaction time measurements in terms of data quantity and time coverage. The graph of Fig. 6, or preferably the TICT portion of Fig. 8, can help the user decide on his best sleep schedule and best day planning for peak performance.

[0081] After a significant number of days of data collected by the monitoring application 110, a graph such as the one in Fig. 7 (illustrating typical“reaction times” but would be applicable to TICTs since they are correlated as shown in Figs. 8-9) can be shown to the user. This graph (preferably the TICT portion of Fig. 9) may be prepared either on the client-side or on the server-side of the application. The graph shows, for each day, a pair of points: 1 ) one point with the go-to-bed timestamp (x-axis) and average alertness level (i.e. , the average intercharacter time interval) for the following day, and 2) one corresponding point with the waking-up timestamp (x-axis) and average alertness level (i.e., the average intercharacter time interval) for that day. The graph shown on Fig. 7 shows maximal alertness levels following particular pairs of bedtime hours.

[0082] The graph shown on Fig. 7, considering the correlation between reaction time and texting speed, would thus show the relationship between bedtime schedule and texting speed, allowing the user to see what bedtimes and out of bed times are related to the fastest texting speed during the day following the bedtime.

[0083] According to an embodiment, the monitoring application 110 is able to relate sleeping hours to the alertness level in order to make bedtime propositions to the user. Optionally, the monitoring application 110 can prompt the user by triggering an alert on the mobile electronic device which encourages the user to go to bed now or in a given period to maximize their average alertness level for the next day. In a more specific embodiment, the bedtime hour proposition can maximize the average alertness level, as inferred from the history of texting speed, for a specific time window of the next day. This time period for future alertness maximization can be inputted by the user in the application, for example by asking that the proposed bedtime maximizes the alertness level at this time of day for the day after; or by having the monitoring application 110 query calendar information from the calendar application of the mobile electronic device, i.e. , propose a bedtime that would maximize the future alertness level for a “meeting” or “presentation” that are provided in the calendar. In other words, the application on the electronic device can determine, for a given period on a future day, which sleeping information (bedtime and out-of-bed time), applied to a sleeping period preceding said future day (e.g., the night before), would induce the higher level of alertness (i.e., shorter average of time TICTs) for the given period in the future day, and outputting a result of said determining by showing the result on the user interface, or by making a visual or auditive alert or sending the result to another computer. [0084] It should be noted that although alertness levels were presented as average time intervals between relevant character typing events on a touchscreen, other measures can be used to present the alertness level, such as a texting rate (also known as speed or frequency), which is the number of typing events per period (e.g., characters per minute or“cpm”), or any other indicator which depends on the TICT while being presented in a way that the user can comprehend.

Example:

[0085] During texting speed calibration, an individual texts a few lines on his mobile device. Most of his texting is represented by continuous texting, but there are a few very brief pauses for thinking, a few mistakes with corrections and some special character search requiring an alternate keyboard change causing texting delays.

[0086] The total number of characters in this example (calibration) including spaces is 100. Of these, the monitoring determines that 80 characters are written with intercharacter time between 220ms and 310ms, and that no intercharacter time are present between 310 and 540ms. The balance of intercharacter times are irregularly dispersed without a specific pattern between 540ms and 3245ms.

[0087] The monitoring application can determine the 300 ms window starting at a X50 point (150 to 450ms, 250 to 550ms, 350 to 650 ms etc.) in which most intercharacter times are present. In this example, the 80 characters with an intercharacter time between 150 and 450ms will be retained for average texting speed. The average texting speed from these results, considering that results are not evenly distributed between 220 and 310 ms and more in the lower range of that window for this individual, is calculated as 252ms.

[0088] Average TICT (intercharacter time) for that individual is calculated to be 252 ms. The window of intercharacter times that will from this point in time be retained to measure texting speed:

252ms x 50% = 126ms 252ms ± 126ms = 126ms and 382ms [0089] From this point, when the texting speed measure is activated in the mobile application, for every hour window starting at :30 (8:30 to 9:30, 9:30 to 10:30, etc.), an average texting speed will be measured from all intercharacter times between 126 and 382ms.

[0090] A lower threshold was discussed as being computed as 50% less than the average TICT; however, other values may prove to be better after calibrating the model, such as 20%, 30%, 40%, 60% or 70% less than the average TICT. Similarly, and independent from the lower threshold, other values of the upper threshold may prove to be better after calibrating the model, such as 20%, 30%, 40%, 60% or 70% more than the average TICT (in addition to the 50% more than the average TICT mentioned above). For example, one can set the lower threshold at a range between 20% and 70%, or preferably between 40% and 60%, under the average time interval and set the upper threshold at a range between 20% and 70%, or preferably between 40% and 60%, above the average time interval.

[0091] The average value of a TICT used to determine the threshold can be computed using all measured values, or advantageously by removing outliers, which are the values which are away from the mean by a large value (predetermined threshold, such as three times the standard deviation, but this threshold can be adjusted). Long pauses would be outliers that are removed as they are not relevant to determine the relative thresholds for the exclusion range of TICTs. This makes the computation of the mean an iterative process, and the average value is less sensitive to the presence of extreme values, which are not representative of the level of alertness, as discussed above. Otherwise, the average used for the thresholds can use the median instead of the mean, since the median is less sensitive to the presence of extreme values that can arise when the user makes long pauses between typing events. In any case, the effect of such long pauses should be minimized when computing an effective average value for threshold determination. [0092] If the thresholds are fixed instead of being relative to an average, the thresholds can be set, for example, between 150ms and 220ms, for the lower threshold, and between 300ms and 500ms, for the upper threshold.

[0093] These are the lower threshold and upper threshold outside of which a time interval is considered to involve other mental activities (the large deviation in TICT is considered not related to alertness); within this threshold the time intervals are considered to be between characters of a sequence written as an automatism, with no other mental task involved (i.e. , no distraction), and thus relevant for assessing alertness levels (the TICT values are related to alertness).

[0094] The hourly averages can thus be calculated and shown on a graph to assess best times of day in terms of alertness. Daily averages can also be shown in comparison with sleeping hours to identify the sleeping hours (going to bed and waking up) that optimize alertness levels. In the present context, the assessment typically is the arithmetic average (i.e., mean) of the TICTs (those not excluded and thus being indicative of alertness levels) over a period (typically between 0.5 and 2 hours). According to another embodiment, the assessment typically is the arithmetic average (i.e., mean) of the TICTs (those not excluded and thus being indicative of alertness levels) over a period of less than 10 seconds.

[0095] Other values or indicators can be derived from the arithmetic average and serve as the assessment instead. Otherwise, other measurements of the center of the distribution (e.g., median instead of mean) can be used for assessment. Such assessment of the alertness level is designed as a replacement (or proxy) for the more academically-accepted version of the assessment of alertness level which is typically (in the prior art) determined through a reaction time test.

[0096] Although the method is described above in reference with a mobile device 10, it can alternatively, or complementarily, be implemented on an actual, material keyboard (i.e., the object used as a peripheral), typically used with a desktop or a laptop computer and which will comprise the monitoring application installed thereon or being run thereon. The computer can thus monitor keyboard pressing events made when the person is typing. A specific calibration would need to be made since TICTs will reasonably be different when the same person is typing on this other device. A single account for the monitoring application can be used to gather data collected on different devices (each one having their own calibration and comparison scale) to ensure maximal data collection coverage over the day (i.e. , using the different devices used by the person to type/text characters).

[0097] Fig. 10 is a flowchart illustrating a method for assessing a level of alertness, according to an embodiment, comprising the following steps:

- step 1010: detecting user typing events on an electronic device;

- step 1020: measuring a time interval between each consecutive one of the user typing events;

- step 1030: excluding all user typing events which correspond to a time interval outside an inclusion range, where the inclusion range characterizes time intervals related to alertness;

- step 1040: producing an assessment of an alertness level using time intervals between user typing events which were not excluded.

[0098] While preferred embodiments have been described above and illustrated in the accompanying drawings, it will be evident to those skilled in the art that modifications may be made without departing from this disclosure. Such modifications are considered as possible variants comprised in the scope of the disclosure.

Claims

CLAIMS:

1 . A method for measuring an alertness level of a user, the method comprising:

- detecting user typing events on an electronic device;

2. The method of claim 1 , wherein detecting user typing events on an electronic device comprises detecting user typing events on a mobile electronic device.

3. The method of claim 2, wherein the steps of the method uses, in addition to user typing events, user screen touches performed outside of a keyboard of the mobile electronic device.

4. The method of claim 2, further comprising providing a monitoring application on the mobile electronic device, wherein detecting user typing events is performed by the monitoring application.

5. The method of claim 4, wherein detecting user typing events is performed when the monitoring application is not open and is performed for user typing events occurring in an application other than the monitoring application.

6. The method of claim 5, wherein detecting user typing events comprises having the monitoring application monitor only typing performed on a keyboard of the mobile electronic device.

7. The method of claim 6, further comprising installing the monitoring application on the mobile electronic device, said installing comprising installing a custom keyboard for use in the application other than the monitoring application, wherein detecting user typing events comprises detecting user typing events performed on the custom keyboard.

8. The method of claim 1 , wherein detecting user typing events on an electronic device comprises detecting user typing events on a keyboard peripheral.

9. The method of claim 1 , further comprising, prior to excluding, determining the inclusion range based on a calibration sequence of typing events, comprising determining a lower threshold and an upper threshold for the time interval.

10. The method of claim 9, wherein determining a lower threshold and an upper threshold comprises determining the lower threshold and the upper threshold relative to an average time interval in the calibration sequence of typing events, the average time interval being computed as one of: a mean time interval of all typing events in the calibration sequence of typing events, a mean of the time interval of the typing events in the calibration sequence of typing events excluding outliers, a median time interval of all typing events in the calibration sequence of typing events, and a median of the time interval of the typing events in the calibration sequence of typing events excluding outliers.

1 1 . The method of claim 10, wherein determining a lower threshold and an upper threshold relative to an average time interval comprises setting the lower threshold at a range between 20% and 70% under the average time interval and setting the upper threshold at a range between 20% and 70% above the average time interval.

12. The method of claim 1 1 , wherein determining a lower threshold and an upper threshold relative to an average time interval comprises setting the lower threshold at a range between 40% and 60% under the average time interval and setting the upper threshold at a range between 40% and 60% above the average time interval.

13. The method of claim 9, wherein determining a lower threshold and an upper threshold comprises setting the lower threshold between 100 and 220ms and setting the upper threshold between 300 and 500ms.

14. The method of claim 1 , further comprising recording, within the monitoring application, sleeping information of the user.

15. The method of claim 14, wherein recording the sleeping information of the user is performed by the user entering the sleeping information in a graphical user interface of the monitoring application.

16. The method of claim 14, wherein recording the sleeping information of the user is performed by the monitoring application on a mobile electronic device detecting events in the environment of the mobile electronic device indicative of the sleeping information.

17. The method of claim 1 , wherein producing the assessment of the alertness level comprises averaging the time intervals between user typing events which were not excluded over a period to produce the assessment of the alertness level which is an average of the time intervals, not excluded, for the period.

18. The method of claim 17, further comprising identifying sleeping information of the user which contribute to higher alertness level characterized by a smaller average of the time intervals for the period.

19. The method of claim 18, further comprising, for a given period on a future day, determining which sleeping information, applied to a sleeping period preceding said future day, would induce the higher alertness level for the given period in the future day, and outputting a result of said determining.

20. The method of claim 17, wherein the period for producing the assessment of the alertness level is between 0.5 hour and 2 hours and wherein the assessment is performed repeatedly.

21 . The method of claim 17, wherein the period for producing the assessment of the alertness level is less than 10 seconds.

22. An electronic device comprising a memory for storing instructions and a processor in communication with the memory for executing the instructions which cause the electronic device to perform the steps of:

- detecting user typing events occurring in the electronic device;

23. The electronic device of claim 22, wherein the electronic device is a mobile electronic device and the typing events are tapping events on a screen of the mobile electronic device.

24. The electronic device of claim 23, further comprising a monitoring application executable by the processor of the mobile electronic device, wherein detecting user typing events is performed by the monitoring application.

25. The electronic device of claim 24, further comprising an application other than the monitoring application executable by the processor of the mobile electronic device, wherein detecting user typing events is performed when the monitoring application is not open and is performed for user typing events occurring in the application other than the monitoring application.

26. The electronic device of claim 25, further comprising a custom keyboard installed on the mobile electronic device and executable by the processor of the mobile electronic device for use in the application other than the monitoring application, wherein detecting user typing events comprises detecting user typing events performed on the custom keyboard.

27. The electronic device of claim 22, further comprising a keyboard peripheral connected to the electronic device for detecting user typing events thereon.