Facilitating Text Entry on Smartphones with QWERTY Keyboard for Users with Parkinson's Disease
Investigating auto-correct model tweaks to increase user confidence in the context of hand tremors
As the world population’s average age increases, Parkinson’s dis- ease has become a challenge for more and more people. In 2020, the number of Parkinson’s patients reached 10 million. Parkinson’s disease is a long-term nervous system disorder that mainly affects the motor system. The most common symptoms are the "pill-rolling" hand tremor (between 4 – 6 hertz) and muscle rigidity/stifness. As a result, Parkinson’s patients usually find fine motor movements (e.g., grabbing spoons and pressing buttons) difficult. Interacting with touchscreen devices is a major challenge for users with Parkinson’s disease, specifically, inaccurate input and accidental touches significantly limit their interaction performance and experience. For example, when typing on smart-phone QWERTY keyboards, inexperienced users with the hand tremor could only type 4.7 words per minute (WPM), about 11% of the speed of young adults.
This paper presents and evaluates a smartphone QWERTY keyboard for users with Parkinson’s disease using an elastic probabilistic model. We first conducted a user survey to explore how users input text in a daily scenario, including the most widely used keyboard layout and typing posture. Then we investigated and compared the typing behaviors generated by both Parkinson’s and non-Parkinson’s users. Finally, we proposed an elastic probabilistic model to correct all major types of errors while maintaining direct physical interpretation by incorporating spatial-temporal features. In a second user study, we evaluated the performance versus two baseline models: the basic language model (BLM)and elastic pattern matching (EM). Results showed that our method achieves significantly higher typing speed (22.8 WPM), 26.8% and 14.6% faster than BLM and EM respectively; as well as lower word-level error rate (8.0%), 7.1% (5.5%) lower than BLM (EM) and keystrokes per character (1.06), 7.8% (5.5%) lower than BLM (EM). Finally, users ranked our method to be the best in terms of perceived accuracy, speed, error correction performance, confidence, and overall preference.
This paper presents and evaluates a smartphone QWERTY keyboard for users with Parkinson’s disease using an elastic probabilistic model. We first conducted a user survey to explore how users input text in a daily scenario, including the most widely used keyboard layout and typing posture. Then we investigated and compared the typing behaviors generated by both Parkinson’s and non-Parkinson’s users. Finally, we proposed an elastic probabilistic model to correct all major types of errors while maintaining direct physical interpretation by incorporating spatial-temporal features. In a second user study, we evaluated the performance versus two baseline models: the basic language model (BLM)and elastic pattern matching (EM). Results showed that our method achieves significantly higher typing speed (22.8 WPM), 26.8% and 14.6% faster than BLM and EM respectively; as well as lower word-level error rate (8.0%), 7.1% (5.5%) lower than BLM (EM) and keystrokes per character (1.06), 7.8% (5.5%) lower than BLM (EM). Finally, users ranked our method to be the best in terms of perceived accuracy, speed, error correction performance, confidence, and overall preference.
|
Citation
Yuntao Wang, Ao Yu, Xin Yi, Yuanwei Zhang, Ishan Chatterjee, Shwetak Patel, and Yuanchun Shi. 2021. Facilitating Text Entry on Smartphones with QWERTY Keyboard for Users with Parkinson’s Disease. Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems. Association for Computing Machinery, New York, NY, USA, Article 735, 1–12. DOI:https://doi.org/10.1145/3411764.3445352 |
Short Video:
|
Conference Presentation (by Yuntao):
|
|
|
© Ishan Chatterjee 2020