Validation of smartphone step count algorithm used in STARFISH smartphone application.

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Rehabilitation Science

Background:

• Smartphone sensors offer a largely untapped resource for remote patient monitoring and rehabilitation.
• Existing rehabilitation technologies often lack the accessibility and widespread adoption of smartphones.

Purpose of the Study:

• To validate the accuracy of the step count algorithm within the STARFISH smartphone application.
• To assess the performance of the STARFISH application across various body-worn positions and walking speeds.

Main Methods:

• Twenty-two healthy adults performed treadmill walking at four distinct speeds (0.44–1.33 m·s⁻¹).
• Step counts from the STARFISH app on Samsung Galaxy S3 smartphones were compared against an activPAL™ device.
• Data were collected with smartphones placed in four different positions: belt carry case, pocket, handbag/shirt pocket, and upper arm strap.

Main Results:

• The STARFISH application's step count accuracy, measured by Level of Agreement (LOA), improved significantly with increased walking speed.
• At the highest speed (1.33 m·s⁻¹), LOA ranged from 8% (shirt pocket) to 53% (belt carry case).
• Optimal accuracy was observed at 1.33 m·s⁻¹ when the smartphone was in a shirt pocket or on an upper arm strap.

Conclusions:

• The STARFISH smartphone application demonstrates valid step count measurements for rehabilitation purposes.
• Accuracy is particularly high at walking speeds of 0.9 m·s⁻¹ and above, and in specific body positions like shirt pockets or upper arm straps.