providing high quality detection by integrating several
optimal threshold-based mechanisms using the force on
the phone and tilt calculations. Through data
classification, optimal thresholds were identified to
effectively distinguish between emergencies (seizures,
falls) and normal activities with sensitivity over 95% and
specificity over 90%. Extensive and long term testing
shows less than 1 false positive per day (with normal use
of phone)1. Majority of the false positives were caused
by high activity (e.g., jogging, dancing). The algorithms
run in the background and were repeatedly refined to
achieve power-awareness with ~5% power consumption
per day. Seizario uses only the tri-axial accelerometer
sensor found in all smartphones making the app
compatible with affordable/lower-end phones. Hence, all
the engineering design goals are met under the given
limited smartphone power and computation resources.