GB2546737A

GB2546737A - Method and device for detecting a smoking gesture

Info

Publication number: GB2546737A
Application number: GB1601342.7A
Authority: GB
Inventors: Skinner Andrew; Stone Christopher; Doughty Hazel; Munafo Marcus
Original assignee: University of Bristol
Current assignee: University of Bristol
Priority date: 2016-01-26
Filing date: 2016-01-26
Publication date: 2017-08-02
Also published as: WO2017129946A1; GB201601342D0

Abstract

A method 40 for detecting smoking gestures has the steps of detecting an initial phase of a smoking gesture when measurement data from a gyroscope 44 exceeds both a first and second threshold 46, subsequently detecting a stationary phase when data from an accelerometer 54 corresponds to a reference dataset 56 and then detecting a final phase when the measurement data from the gyroscope 60 exceeds the first threshold 62. Detection of the initial phase may include determining whether the measurement data falls between the first and second thresholds for a predetermined period of time 52. A smoking event may be detected by counting the number of smoking gestures and determining if this exceeds a threshold. An alert may be issued if a smoking event or a smoking gesture is detected. A device 10 is also claimed and includes a gyroscope 18, an accelerometer 20, a processor 16 and a memory 22 which includes instructions which can be executed by the processor to perform the preceding method.

Description

METHOD AND DEVICE FOR DETECTING A SMOKING GESTURE

Technical Field

The present application relates to a method and device for detecting a smoking gesture, that is to say physical movements of the human body that take place during the act of smoking a cigarette or the like.

Background to the Invention

Smoking of cigarettes and other tobacco products remains a major health problem throughout the world, despite efforts, such as the use of graphic images on tobacco packaging and the prohibition of smoking in public places, to discourage and de-normalise smoking. It is estimated that six million people per year die from smoking related causes. It is well known that nicotine in tobacco products is highly addictive, and this addictive quality makes it very difficult for many smokers to stop. A number of programs have been developed which show that smokers can be helped to stop smoking if intervention occurs at opportune moments. A key factor in this is detecting smoking in real world settings. If this can be done accurately, continuous evaluation of a smoker's smoking pattern over time can be combined with the contextual factors that influence smoking behaviour, such as environment, in order to inform timely interventions.

Techniques and devices have been proposed to provide such interventions. For example, in their paper "RisQ: Recognizing Smoking Gestures with Inertial Sensors on a Wristband", Parate, Chiu, Chadowitz, Ganesan and Kalogerakis describe a wristband containing a 9-axis inertial measurement unit which is operative to capture changes in the orientation of a person's arm and a machine learning pipeline that processes this data to detect accurately smoking gestures and sessions in real-time. One disadvantage of this approach, however, is that it requires an inertial measurement unit that incorporates an accelerometer, a gyroscope and a compass. This combination of sensors is not widely available in a single wearable device, and so widespread adoption of the approach disclosed by Parate et al is limited by the number of devices that are able to provide the required functionality. Rather than relying on commercially available devices, an alternative would be to develop a custom wearable device supporting the required sensor functionality, which would have to be worn routinely by a smoker trying to stop. As will be appreciated, however, any method that requires a dedicated additional device with which a user must actively engage risks failure due to the inconvenience and possible embarrassment to the user of wearing the device. Moreover, as the inertial measurement unit including an accelerometer, a gyroscope and a compass is not widely available, additional cost and time are required to develop such a unit.

Accordingly, a need exists for a method for providing timely interventions to a smoker that does not require a combination of accelerometer, gyroscope and compass.

Summary of Invention

According to a first aspect of the present invention there is provided a method, performed by a device comprising an accelerometer and a gyroscope, for detecting a smoking gesture, the method comprising: detecting, based on measurement data from the gyroscope, an initial phase of the smoking gesture, in which a user's hand moves towards the user's mouth; if the initial phase is detected, subsequently detecting, based on measurement data from the accelerometer, a stationary phase of the smoking gesture, in which the user's hand is in a smoking position; and if the stationary phase is detected, subsequently detecting, based on data from the gyroscope, a final phase of the smoking gesture, in which the user's hand moves away from the user's mouth, wherein: detecting the initial phase comprises comparing the measurement data from the gyroscope to a first threshold and to a second threshold, and, if the measurement data exceeds both the first and second thresholds, determining that the initial phase has been detected; detecting the stationary phase comprises comparing the measurement data from the accelerometer to a dataset containing reference accelerometer data for smoking gestures and, if the measurement data corresponds to data in the reference dataset for smoking gestures, determining that the stationary phase has been detected; and detecting the final phase comprises comparing the measurement data from the gyroscope to the first threshold and, if the measurement data exceeds the first threshold, determining that the final phase has been detected.

The method of the present invention can be implemented using commercially available sensors and devices and facilitates rapid and accurate detection of smoking gestures, thereby permitting timely and appropriate interventions to assist a smoker in giving up the habit. Because the method can be implemented in existing devices, there is no need for a dedicated device to be worn by a smoker. Instead, the method can be implemented as an application or program running on an existing device such as a smartwatch that the smoker would normally wear. This facilitates greater compliance with a program for stopping smoking.

Moreover, as the method requires data from only two sensors, the power consumption of a device implementing the method is reduced, in comparison to a device that uses data from three sensors, thereby helping to maximise the available battery life of the device. Detecting the initial phase may further comprise determining whether the measurement data falls between the first and second thresholds for a predetermined period of time.

This helps to avoid false positives, in which movements relating to activities other than smoking are misinterpreted as smoking related movements.

The predetermined period of time may be between 0.3 and 0.7 seconds, for example.

Detecting the initial phase may further comprise detecting the end of the movement of the user's hand towards the user's mouth.

Again, this helps to avoid false positives.

Detecting the end of the movement of the user's hand towards the user's mouth may comprise comparing the measurement data from the gyroscope to the first threshold and, if the measurement fails to meet the first threshold, determining that the end of the movement of the user's hand towards the user's mouth has been detected.

Detecting the stationary phase may further comprise determining whether the measurement data corresponds to data in the reference dataset for smoking gestures for a predetermined period of time.

The predetermined time may be between 0.6 second and 5.5 seconds, for example.

Detecting the final phase may further comprise comparing the measurement data from the gyroscope to the second threshold and, if the measurement data exceeds both the first threshold and the second threshold, determining that the final phase has been detected.

The method may further comprise: counting the number of completed smoking gestures detected; comparing the number of completed smoking gestures detected to a threshold; and if the number of completed smoking events detected meets the threshold, determining that a smoking event has been detected.

The threshold may be 6, for example.

The method may further comprise issuing an alert if a smoking event is detected.

Alternatively, the method may further comprise issuing an alert if a completed smoking gesture is detected.

Thus, a smoker can be alerted if a complete smoking event (e.g. smoking a whole cigarette) or a single smoking gesture (e.g. taking a single "drag" of a cigarette) has been detected. These alerts can help to discourage the smoker from continuing to smoke.

The alert may comprise one or more of: a vibration; a sound; a written message; and an audible message.

The measurement data from the gyroscope may comprise a sum of absolute angular velocity values, in radians per second, about x, y and z axes.

The first threshold may be 0.6 radians per second and the second threshold may be 1.5 radians per second, for example.

According to a second aspect of the invention there is provided a device for detecting smoking gestures, the device comprising: an accelerometer; a gyroscope; a processor; and a memory, wherein the memory includes instructions which, when executed by the processor, cause the device to perform the method of any one of the first aspect.

According to a second aspect of the invention there is provided a computer program which, when executed by a processor, performs the method of the first aspect.

Brief Description of the Drawings

Embodiments of the invention will now be described, strictly by way of example only, with reference to the accompanying drawings, of which:

Figure 1 is a schematic representation of a wrist-worn device for detecting a smoking gesture; and

Figure 2 is a flow diagram illustrating steps of a method for detecting a smoking gesture. Description of the Embodiments

Figure 1 is a schematic representation of a wrist-worn device for detecting smoking gestures. In this context, a smoking gesture is a series of actions that occur during smoking of a cigarette, which may include: an initial phase or action in which a cigarette or the like, which is held in a smoker's hand, is raised towards the smoker's mouth; a stationary phase in which the smoker's hand is substantially stationary as the smoker "drags" (i.e. sucks or draws) on the cigarette; and a final phase in which the smoker moves the cigarette away from the mouth.

The device, shown generally at 10, comprises a main body 12 to which is connected a strap 14, which permits the device 10 to be worn on the wrist. The main body 12 houses a processor 16, a gyroscope 18, an accelerometer 20, memory 22, a display 24 and one or more output devices 26 which may include, for example, a vibrator, a speaker, a light and/or other actuators or output devices suitable for drawing a wearer's attention to the device 10.

The processor 16 is configured to access instructions stored in the memory 22 which, when implemented by the processor 16, cause the device 10 to perform a method for detecting smoking gestures, as will be described below.

It will be appreciated that many commercially available smartwatch devices include the combination of processor, memory, sensors and output devices described above and shown in Figure 1. Thus, while the device of the present invention may be implemented in a dedicated device, it is envisaged that in practice the device of the present invention will be implemented as a commercially available smartwatch device on which is installed and executed an application or program for performing the novel method described below with reference to Figure 2.

The gyroscope 18 is connected to the processor 16 and provides output signals representing the rate of rotation about three orthogonal (x, y and z) axes, which can be used to detect movement of the device 10. The accelerometer 20 is also connected to the processor 16 and provides output signals that can be used to determine the orientation of the device 10 relative to the ground. The gyroscope 18 and the accelerometer 20 periodically (e.g. every 225 milliseconds) provide output signals to the processor 16, and these output signals are used by the processor 16 to trigger a determination as to whether a smoking gesture is being performed, as will be explained below.

The display 24 provides visual outputs to the wearer of the device, including, for example, text and/or graphical notifications that a smoking gesture has been detected. The one or more output devices 26 provide additional outputs to the wearer of the device, for example in the form of vibrations, illuminations or sounds. A method for detecting smoking gestures made by a wearer of the device 10 will now be described with reference to the flow chart of Figure 2.

The method (shown generally at 40) starts at step 42. In a first step 44 the processor 16 receives data from the gyroscope 18 in the form of values, in radians per second, for the rate of rotation about the three orthogonal (x, y and z) axes. This data may be provided by the gyroscope 16 either as one of the periodic output signals discussed above, or may be provided in response to a request issued by the processor 16.

At step 44 the processor 16 uses the data received from the gyroscope 18 to determine whether the wearer's hand is moving. This step is part of a process by which the processor 16 detects an initial phase of a smoking gesture, in which the wearer's hand (holding a cigarette), moves towards the wearer's mouth prior to the smoker "dragging" (i.e. sucking or drawing) on the cigarette.

Laboratory data shows that in this initial phase of a smoking gesture, as the smoker's hand moves in a smooth motion from a resting position towards the mouth the smoker's forearm rotates.

In order to detect this motion of the hand towards the mouth, and the attendant rotation of the forearm, the processor 16 sums the absolute x, y and z values received from the gyroscope 18, and compares the resulting value to a first, lower, threshold, above which it can be inferred that the wearer's hand is not stationary, and to a second, higher, threshold, above which it can be inferred that the movement of the hand is a deliberate movement of the hand towards the mouth. Laboratory data shows that a suitable value for the first, lower, threshold is 0.6 radians/second and that a suitable value for the second, higher, threshold, is 1.5 radians/second.

If the sum of the absolute x, y and z values does not meet the first, lower, threshold, then it can be inferred that the wearer's hand is not moving, and so there is no need for the processor 16 to perform a comparison with the second, higher, threshold. The method therefore returns to step 44, to await new gyroscope data. However, if the sum of the x, y and z values meets or exceeds the first, lower threshold, it can be inferred that the hand is not stationary, and the processor 16 must therefore must perform a comparison of the sum of the x, y and z values with the second, higher, threshold.

If the sum of the x, y and z values does not meet the second, higher, threshold, it can be inferred that the detected movement is not a deliberate movement of the hand towards the mouth, and so no further action is required until new data is received by the processor 16 from the gyroscope 18 (at step 44). Finally, if the sum of the x, y and z values meets or exceeds the second, higher, threshold, it can be inferred that the detected movement is a deliberate movement of the hand towards the mouth, and so the processor 16 stores the results of the comparison against the first and second thresholds in a database held in the memory 22 as part of a database entry of constituent parts of a smoking gesture, and the method moves on to step 48.

In step 48, the processor 16 receives new data from the gyroscope 18, again either as one of the periodic output signals discussed above, or in response to a request issued by the processor 16.

At step 50, the processor 16 analyses the received data to determine whether the wearer's hand has stopped moving, which occurs at the end of the initial phase of the smoking gesture. To do this, the processor 16 again sums the absolute x, y and z values received from the gyroscope 18, and compares the resulting value to the first, lower, threshold.

If the sum of the absolute x, y and z values meets or exceeds the first, lower, threshold, then it can be inferred that the wearer's hand has not stopped moving, and thus the method returns to step 48 to obtain new gyroscope data.

On the other hand, if the sum of the absolute x, y and z values fails to meet the first, lower, threshold, then it can be inferred that the wearer's hand has stopped moving, and the processor 16 records this in the database held in the memory 22.

The method then moves on to step 52, in which a determination is made as to whether the duration of the detected movement of the wearer's hand was long enough to represent a movement of the hand towards the face during an initial phase of a smoking gesture.

To do this, the processor 16 compares the duration of the detected movement to a first, lower, threshold duration. Laboratory data shows that a movement of the hand lasting for a short duration (e.g. less than around 0.3 seconds) is unlikely to correspond to a deliberate movement of a hand towards the mouth prior to sucking or drawing on the cigarette. Thus, this first, lower, threshold duration may be 0.3 seconds.

If the duration of the detected movement fails to meet the first, lower, threshold duration, it can be inferred that the detected movement is not part of a smoking gesture and no further action is required. The method then returns to step 44, to receive new data from the gyroscope 18 which may be indicative of a movement of the wearer's hand towards the face.

If the duration of the detected movement meets or exceeds the first, lower, threshold duration, the processor 16 then compares the duration of the detected movement to a second, higher, threshold. Laboratory data shows that a movement of the hand lasting for a duration of more than around 0.7 seconds is likely to correspond to movement of the hand to perform a task other than raising a cigarette to the mouth. Accordingly, if the duration of the detected movement meets or exceeds the second, higher, threshold duration (which may be, for example, 0.7 seconds), then it can be inferred that the movement is for some purpose other than raising a cigarette to the mouth, and no further action is required. The method then returns to step 44, to receive new data from the gyroscope 18 which may be indicative of a movement of the wearer's hand towards the face.

If the duration of the detected movement meets or exceeds the first threshold duration and also fails to meet the second, higher, threshold duration, then it can be inferred that the detected movement corresponds to a movement to raise a cigarette to the mouth as part of a smoking gesture. Thus, the processor 16 stores the results of the comparison against the first and second threshold durations in the database held in the memory 22.

If the database contains entries indicating that the detected movement represents a deliberate movement of the hand towards the mouth, as determined in step 46, and that the movement has stopped, as determined in step 50, and that the duration of the movement was sufficient, as determined in step 52, the processor 16 may determine that the initial phase of the smoking gesture has been completed, and may update the database accordingly. Alternatively, the processor 16 may update the database if only two or fewer of these three steps generate a positive result. For example, if movement of the hand towards the mouth is detected in step 46, and the duration of this movement is sufficiently long, as determined in step 52, then the processor 16 may update the database to indicate that the initial phase of the smoking gesture has been detected. Similarly, the processor 16 may update the database if, for example, movement of the hand towards the mouth is detected in step 46. Thus, in some implementations it is not necessary to perform all three of steps 46, 50 and 52 in order to determine that the initial phase of the smoking gesture has been detected.

Following a determination by the processor 16 that the initial phase of the smoking gesture has been detected, the processor 16 proceeds to determine whether a stationary phase of the smoking gesture, which corresponds to a period in which the smoker's hand is substantially stationary as the smoker sucks or draws on the cigarette, has been detected.

The method thus moves to step 54, in which accelerometer data is received by the processor 16 from the accelerometer 20. The accelerometer data may be received in response to a specific request issued by the processor 16, or may be received as a result of the periodic transmission of data by the accelerometer 20 to the processor 16.

At step 56 the processor 16 analyses the data received from the accelerometer 20 to determine whether the wearer's hand is in a smoking position. To do this the processor 16 compares the data (x, y and z values) received from the accelerometer 20 to data of a reference dataset containing reference accelerometer data (x, y and z values) for smoking and non-smoking gestures. This comparison may be performed, for example, by inputting the received accelerometer data into a decision tree trained on the reference dataset, which contains accelerometer data obtained during the stationary phase of smoking gestures in a number of smoking "styles" performed in a variety of sitting and standing positions. The reference dataset also contains accelerometer data from activities other than smoking such as eating and drinking, to facilitate discrimination between smoking movements and movements that are attributable to other activities.

If the received accelerometer data does not correspond sufficiently to data of the reference dataset that is characteristic of a smoking position for a determination to be made that the wearer's hand is in a smoking position, or corresponds to a data of the reference dataset that is characteristic of a non-smoking activity, then the processor 16 takes no further action and the method returns to step 44, in which new gyroscope data is received by the processor 16.

If the received accelerometer data corresponds sufficiently to data of the reference dataset that is characteristic of a smoking position for a determination to be made that the wearer's hand is in a smoking position, the processor 16 updates the database (at step 58) to indicate that the wearer has started smoking and that a drag on the cigarette has started.

The processor 16 must then determine when the drag on the cigarette ends. This is achieved by obtaining new gyroscope data, at step 60, which is analysed at step 62 to determine if there is significant movement of the hand.

In a typical smoking gesture the stationary phase in which the smoker drags or draws on the cigarette is followed by a final phase in which the smoker moves the cigarette away from the mouth. In step 62 the processor therefore analyses the gyroscope data received at step 60 to determine whether significant movement of the hand is taking place, which would be indicative that this final phase of the smoking gesture had been commenced or completed.

Thus, as in step 46, the processor 16 sums the absolute x, y and z values of the received gyroscope data and compares the result of the summation to a first, lower, threshold (which, as before, may be 0.3 radians per second) to determine if the wearer's hand is moving. In some implementations, a determination that the wearer's hand is no longer stationary is sufficient to infer that the wearer has removed the cigarette from the mouth and therefore the drag on the cigarette has ended. In other implementations, however, a second comparison step may be performed by the processor 16, to compare the summed gyroscope value to a second, higher, threshold (which, as in step 46, may be 0.7 radians per second) to determine if the detected movement is a deliberate movement of the wearer's hand away from the mouth.

In either event, if the result of the comparison at step 62 is a determination that significant movement of the hand is not taking place, the method returns to step 60, in which new gyroscope data is received by the processor 16.

If the result of the comparison at step 62 is a determination that significant movement of the hand is taking place, the method moves to step 64, in which the processor 16 records in the database that the stationary phase has finished.

The processor 16 then determines whether the detected stationary phase corresponds to a drag on a cigarette, by comparing the duration of the detected stationary phase to a first, lower threshold duration and a second, higher, threshold duration.

Laboratory data shows that a typical drag on a cigarette lasts between 0.6 seconds and 5.5 seconds. Thus, the first, lower, threshold duration may be 0.6 seconds and the second, higher, threshold duration may be 5.5 seconds. If the duration of the stationary phase fails to meet the first, lower, duration or exceeds the second, higher, threshold, then it can be inferred that the detected stationary phase does not correspond to a drag on a cigarette. However, if the duration of the stationary phase meets or exceeds the first, lower threshold and also fails to meet the second, higher, threshold, it can be inferred that the detected stationary phase corresponds to a drag on a cigarette. In this case, the processor 16 updates the database (at stage 64) to record that the drag has finished.

Once a complete smoking gesture entry (including the initial phase, stationary phase and final phase) has been completed in the database, the processor counts the number of completed smoking gesture entries that have been completed in a predetermined time period, for example a period starting when a first drag is detected and ending 80 seconds after the last drag that has been detected, to determine whether the wearer is or has been smoking, rather than performing some other action such as eating or drinking. Thus, in the example above, where no new drag has been detected for a period of 80 seconds after finishing a drag, the last detected drag is deemed to be the final drag and the wearer is deemed no longer to be smoking.

If the number of smoking gesture entries recorded in the database meets or exceeds a threshold in the predetermined time period, the processor 16 may determine that the wearer is smoking, and the device 10 may issue an alert or notification to the wearer, via the display 24 and/or one or more of the output devices 26, to discourage the wearer from continuing to smoke. The notification or alert may take many forms, for example a vibration, a sound, a written message displayed on the display 24 or an audible message output by a speaker of the device 10.

Laboratory data shows that a typical smoking event will involve a minimum of six drags on a cigarette over a period commencing on detection of the first drag and finishing when no further smoking activity is detected after the last drag. Thus, the threshold to which the number of completed smoking gesture entries is compared may be set at six.

Alternatively, the device 10 may issue an alert or notification to the wearer as soon as a single completed smoking gesture has been detected, to discourage the wearer from continuing to smoke.

Additionally or alternatively, if no new smoking gesture is detected within a predetermined period of time (e.g. 80 seconds) from the end of the previous detected smoking gesture, then it can be inferred that the wearer has finished smoking, and the processor 16 therefore updates the database to show that the current smoking event is finished. The device 10 may issue an alert or notification to the wearer when the processor 16 determines that the smoking event has finished in this way.

It will be appreciated that the method 40, or the steps 44, 48, 54 and 60 may be triggered by movements other than smoking gestures which cause the gyroscope 18 data values or the accelerometer 20 data values to change, or by the periodic output of data by the gyroscope 18 and accelerometer 20. However, it will also be appreciated that the method 40 includes numerous steps that reduce the likelihood that a detected movement will be incorrectly interpreted as a smoking gesture or part of a smoking gesture. In particular, the detection of the three distinct phases of a smoking gesture and the requirement for a number of smoking gestures to be completed before a notification is issued to the wearer of the device 10 greatly reduces the risk of false notifications being issued.

The method and device described herein permit the detection of smoking gestures, and facilitates timely and appropriate interventions at the point at which a smoking event (e.g. smoking a cigarette) or even a single smoking gesture (e.g. a single drag on a cigarette) has been detected. These interventions can assist a smoker's attempts to stop smoking. The device and method can be implemented using commercially available sensors and devices, thus reducing the time and cost involved in developing and distributing the device and method to smokers.

Claims

1. A method, performed by a device comprising an accelerometer and a gyroscope, for detecting a smoking gesture, the method comprising: detecting, based on measurement data from the gyroscope, an initial phase of the smoking gesture, in which a user's hand moves towards the user's mouth; if the initial phase is detected, subsequently detecting, based on measurement data from the accelerometer, a stationary phase of the smoking gesture, in which the user's hand is in a smoking position; and if the stationary phase is detected, subsequently detecting, based on data from the gyroscope, a final phase of the smoking gesture, in which the user's hand moves away from the user's mouth, wherein: detecting the initial phase comprises comparing the measurement data from the gyroscope to a first threshold and to a second threshold, and, if the measurement data exceeds both the first and second thresholds, determining that the initial phase has been detected; detecting the stationary phase comprises comparing the measurement data from the accelerometer to a dataset containing reference accelerometer data for smoking gestures and, if the measurement data corresponds to data in the reference dataset for smoking gestures, determining that the stationary phase has been detected; and detecting the final phase comprises comparing the measurement data from the gyroscope to the first threshold and, if the measurement data exceeds the first threshold, determining that the final phase has been detected.

2. A method according to claim 1 wherein detecting the initial phase further comprises determining whether the measurement data falls between the first and second thresholds for a predetermined period of time.

3. A method according to claim 2 wherein the predetermined period of time is between 0.3 and 0.7 seconds.

4. A method according to any one of the preceding claims wherein detecting the initial phase further comprises detecting the end of the movement of the user's hand towards the user's mouth.

5. A method according to claim 4 wherein detecting the end of the movement of the user's hand towards the user's mouth comprises comparing the measurement data from the gyroscope to the first threshold and, if the measurement fails to meet the first threshold, determining that the end of the movement of the user's hand towards the user's mouth has been detected.

6. A method according to any one of the preceding claims wherein detecting the stationary phase further comprises determining whether the measurement data corresponds to data in the reference dataset for smoking gestures for a predetermined period of time.

7. A method according to claim 6 wherein the predetermined time is between 0.6 second and 5.5 seconds.

8. A method according to any one of the preceding claims wherein detecting the final phase further comprises comparing the measurement data from the gyroscope to the second threshold and, if the measurement data exceeds both the first threshold and the second threshold, determining that the final phase has been detected.

9. A method according to any one of the preceding claims further comprising: counting the number of completed smoking gestures detected; comparing the number of completed smoking gestures detected to a threshold; and if the number of completed smoking events detected meets the threshold, determining that a smoking event has been detected.

10. A method according to claim 9 wherein the threshold is 6.

11. A method according to claim 9 or claim 10 further comprising issuing an alert if a smoking event is detected.

12. A method according to any one of claims 1 to 8 further comprising issuing an alert if a completed smoking gesture is detected.

13. A method according to claim 11 or 12 wherein the alert comprises one or more of: a vibration; a sound; a written message; and an audible message.

14. A method according to any one of the preceding claims wherein the measurement data from the gyroscope comprises a sum of absolute angular velocity values, in radians per second, about x, y and z axes.

15. A method according to claim 14 wherein the first threshold is 0.6 radians per second and the second threshold is 1.5 radians per second.

16. A device for detecting smoking gestures, the device comprising: an accelerometer; a gyroscope; a processor; and a memory, wherein the memory includes instructions which, when executed by the processor, cause the device to perform the method of any one of the preceding claims.

17. A computer program which, when executed by a processor, performs the method of any one of claims 1-15.