MyMove: Facilitating Older Adults to Collect In-Situ Activity Labels on a Smartwatch with Speech

Document Title

MyMove: Facilitating Older Adults to Collect In-Situ Activity Labels on a Smartwatch with Speech

Document Information

• Research Area: Human-Computer Interaction, Wearable Device Data Collection, Health Technology Assistance for Older Adults
• Keywords: Activity Labels, Older Adults, Smartwatch, Speech Interaction, Experience Sampling Method

Research Background and Problem

• What issues or challenges did the authors identify?

  • Current activity tracking technologies are primarily trained on data from younger populations, failing to accurately reflect the characteristics of older adults (e.g., gait, slower movements), limiting their applicability for supporting older adults' health.
  • Older adults have a low adoption rate of existing activity tracking devices and lack trust in their accuracy.
  • The unique physical activities and daily routines of older adults require a tailored activity tracking technology solution.

• Why is this problem important?

  • Physical activity significantly impacts older adults' health, psychological well-being, and longevity. Activity tracking technology that aligns with their real habits and needs can provide critical support for public health and personalized care.

• Research Motivation and Related Work

  • The motivation lies in developing personalized activity tracking technology that integrates older adults' activity data to support healthy lifestyles.
  • Related work includes studies on activity types for older adults, behavior data collection technologies, and activity labeling through devices, but lacks systems specifically designed and validated for older adults.

Solution

• What methods or solutions did the authors propose?

  • Development of a voice-interactive smartwatch application, "MyMove," to help older adults label activity data in real-time with minimal data collection burden.
  • Design of a simplified data capture flow integrating voice input, combining free reporting and notification-based activity triggers.

• What is innovative about this solution?

  • The first activity labeling tool specifically designed for older adults, combining the flexibility of voice input with the convenience of a smartwatch.
  • Enables users to describe activity types, duration, and intensity in natural language, enhancing the ease and applicability of labeling.
  • Demonstrates the potential for optimizing activity tracking through data complementarity and personalized training models in multi-device environments.

• What are the implementation steps? What key technologies were used?

  • Development of the MyMove application, integrating voice input, notification prompts, and automatic recording of sensor data (e.g., accelerometer, heart rate, step count).
  • Deployment of a 7-day experiment with older adults to collect voice reports and sensor data for system validation.
  • Analysis of the semantics of voice reports, completeness of time annotations, and their correspondence with sensor data, along with evaluation of voice recognition accuracy.

Research Outcomes

• What specific results were achieved?

  • During the experiment, 13 older adult participants submitted 1,224 voice reports covering 29 activity types and diverse life scenarios.
  • All reports were successfully transcribed into text, with low error rates in automatic speech recognition (Microsoft: 4.93%, Google: 8.50%).
  • The study found that time information and activity intensity in reported activities corresponded to low-intensity movements recorded by sensors, such as informal walking.

• What advantages does it have compared to existing solutions?

  • Older participants demonstrated high engagement (an average of 13.45 reports per day, with smartwatch wear time averaging 11.6 hours/day), indicating the acceptability and practicality of this voice labeling approach.
  • Compared to traditional text box input methods, voice input reduces interaction burden while improving the naturalness of expression and completeness of activity semantic information.

• What were the experimental or evaluation results?

  • Data analysis showed that time annotations in voice reports were relatively complete, with 55.25% of activities accurately aligned with time information.
  • Comparison with sensor data revealed that participants' subjective evaluations of activity intensity did not fully align with standard physical activity intensity ranges.

• Limitations and Future Directions

  • Limitations include a small sample size and a participant pool skewed toward highly educated, tech-savvy older adults, not fully representing groups with varying health conditions.
  • Future work will focus on optimizing voice input for participants with visual impairments or motor disabilities and improving recognition of challenging voice samples.
  • Exploring more accurate integration and data analysis in cross-device environments, as well as encouraging self-monitoring through activity feedback, are potential directions for further research.