Physical Activity Evaluation Using a Voice Recognition App: Development and Validation Study

Area of Science:

• Biomedical Engineering
• Health Informatics
• Wearable Technology

Background:

• Traditional physical activity evaluation methods (questionnaires, accelerometers) have limitations.
• Web-based systems offer easier data collection.
• Voice recognition technology can streamline physical activity monitoring without device interaction.

Purpose of the Study:

• To develop a smartphone app utilizing voice recognition for behavior and physical activity recording.
• To validate the app's accuracy by comparing its data with accelerometer measurements.

Main Methods:

• Developed a voice recognition system integrated into a smartphone app.
• 20 participants wore 3-axis accelerometers and used the app for 7 days.
• Exercise intensity was quantified using metabolic equivalents (METs) derived from voice data.

Main Results:

• The app achieved 78.4% automatic conversion of voice data to behavioral words.
• Mean daily METs values from the app (1.64) and accelerometer (1.63) showed no significant difference (P=.57).
• A strong correlation (r=0.830, P<.001) was found between METs measured by the app and accelerometer, with a small mean difference (0.02 METs).

Conclusions:

• The voice recognition app provides a valid and highly correlated measure of physical activity compared to accelerometers.
• This technology enables simultaneous, low-burden physical activity assessment for large groups.
• Further improvements in data classification and user compliance are needed, but voice recognition presents a novel approach to physical activity evaluation.