Kinect-Based Persuasive Method for a Healthy Sitting Posture

Did you ever think of a smart computer screen to alert your unhealthy sitting postures? We have designed such a device: by reading the depth data, the computer screen blurs out when the user is putting his head too close to the screen. Otherwise, the screen works normally. We tested its effectiveness in a real-subject experiment.


Summary

In this paper, we presented an interactive method named as Self-Adaptive Blur (SAB). By altering the blurriness of the screen according to the context, the SAB method was designed to prevent common discomforts of computer users by helping them maintain healthy posture. A functioning prototype was developed using Microsoft Kinect and a desktop client program. The efficacy of our persuasive practice was examined by a series of experiments with the prototype. Objective records demonstrated the effectiveness of the SAB method, while subjective questionnaires and interviews revealed other instructive findings for further improvements.


Background

Discomforts caused indirectly by computer usage threaten the health and performance of computer users. The cause of the most common discomforts, the neck and shoulder pains (NSP), are kyphotic sitting postures, which means the user puts his head excessively forward, and thus exerts extra and prolonged extension and stress to the musculoskeletal system. We have designed a way for the computer to alert its users: by reading the depth data from the sensor, the computer screen blurs out, when an unhealthy sitting posture is detected, otherwise, the screen works normally.

May 2013 - Jan. 2014


Team

2 Industrial Engineering Students, 1 Computer Science Student


My Responsibilities

  • Preliminary idea generation
  • Experiment designs and conduction
  • Experiment data analysis


Process

IDEA GENERATION

Although massage and several other therapeutic methods are effective in releasing pains, their effectiveness of curing are not obvious. Although sever pains need a surgery, they can be prevented from happening. So Prevention is the best solution to save people from endless pains. The persuasive technology seeks to change people attitudes or behaviors via interactive persuasion - feedback at the right time and at appropriate places.

So we thought about if there is a persuasive technology that interactively prevents neck and shoulder pains related to computer use. We called it "smart screen". It should have a context awareness of users' sitting postures, and be capable to analyze and give a proper feedback.

The upper one shows a healthy sitting posture, while the lower one shows an unhealthy one, since he puts his head too forward to his shoulder.

When the computer user stays in a proper range, he could read everything on the screen clearly, otherwise there will be an alert for his unhealthy sitting posture.

Then we came across our system architecture:

  • Input: Kinect as sensor
  • Process: computer program
  • Output: screen sharpness (blurriness)


Its principle is simple. Just as you adjust the length of a lens to make a photo clearer, you could come closer or further to a screen to adjust the sharpness of the screen. In case the computer user stays in a proper range to the screen (not putting his head too forward), everything on the screen becomes clear. Otherwise, it is blurred to the severity of unhealthy posture.


EXPERIMENT DESIGN

After developing the high-fidelity prototype, we would like to conduct an experiment to:

  • Verify the effectiveness of the SAB method and 
  • Find usability issues and improvements


For the experiment design, the sitting posture (quantitative) and subjective ratings (quantitative & qualitative) were chosen to be the dependent variables, while the type of tasks (reading, clicking, or typing) and the inclusion of feedback were varied as independent variables. A 2*3 full-factorial design was performed.

Details of each task could be found from the pictures below.

In this task, the participant was challenged to pick out the one and only different Chinese character from the matrix of other confusing characters. For example in this matrix, only one character is different. The answer is in the last picture.

The Chinese character marked with a green background is the one and only different character in the matrix. The participant have different sets of characters as tasks.

In this task, the participant was challenged to click the circle shown on the screen. Only one circle was shown on the screen at one time at a random place. The more circles they managed to click, the better they did in this task.

In this task, the participant was challenged to enter a passage of text into the given text box. This is to simulate the real context in an office workplace.


EXPERIMENT APPARATUS

A VB.Net-based program was coded on two sides. The participant side provided the given tasks, and the experimenter side functioned to control the experiment sequence. As can be seen from the picture, the experiment was carried out in an ergonomics lab, where the participant and the experimenter sit opposite to each other during the experiment.

During the experiment, the experimenter and the participant sit opposite to each other.

After the experiment, the participant was also shown another feedback type: the red screen. As this participant put his head too forward, he got a red screen as a feedback.

A VB.Net-based program was coded on two sides.

In this console, the experimenter can track all the data. Meanwhile, a realtime green skeleton was drawn over the participant.

On the interface from the experimenter side, real-data, including the skeleton of the participant, was shown to ensure that data was collected properly. And the experimenter was free to alter the experiment sequence according to the plan so as to balance the learning effect.

After the experiment, the participant was also shown with other kinds of feedback alert to obtain more opinion from them. In the picture above, the participant was shown to the "red screen", which means that the screen turns red when an excessive forward head is detected.


RESULT ANALYSIS

The experiment was performed on 30 subjects (11F/19M), among which 24 (9F/15M) were validated due to the inaccuracy of the Kinect sensor. The objective data, as the sitting posture against the type of tasks and the inclusion of feedback, and the subjective data, as the questionnaire (5 general + 3 SAB-specified questions) and preference and improvement in the follow-up experiment, were collected.

alert_level_of_participant_29_Kinect_SAB

The ANOVA showed that, under a significant level of 0,05, the inclusion of feedback has a significant impact on the averaged alert level, which is the extend to which the participant has put his head forwards. The average alert level with SAB is 6.9 while the value without SAB is 14.7. As can be seen from the following graph, the alert level for participant #29 is significantly lowered with the inclusion of SAB.

The questionnaire data also shed some lights on this effect. During the experiment, high perceived mental and physical concentration were paid to tasks, however, mildly low self-awareness of postures were mentioned. The SAB method showed users high awareness of their unhealthy postures with only very low annoying rate reported.


CONCLUSIONS & FUTURE WORK

We proposed the philosophy and design of the SAB method. An experiment was conducted to testify its effectiveness in helping computer users to maintain a healthy sitting posture. However, three challenges exist for future development. First, the feedback mechanism should be more intelligent, so that it is fit into the working context. A rapid change of the screen sharpness should be avoided to improve user experience. Second, long-term effectiveness should be tested with a more deliberate experiment design. Last but not least, ethical issues such as user privacy should be considered in the future research framework.


Learnings & achievements

  • How to code Kinect-based program with VB.Net
  • Experience with the experiment design and ANOVA analysis
  • 1 EI-indexed conference paper published in the AHFE2014


Back to portfolio


Welcome to your comments!


© Chen Liao, Last update on May 19th, 2017