Key innovations of the mPerf Project

mPerf will collect data from a suite of mobile sensors, both body-worn and situated in the workplace. These unobtrusive sensors have already been used in mHealth research to identify biomarkers of stress and other behaviors related to health and wellness.

The sensors include:

For this project, we are enlisting 600 study participants who work in fast-paced, information-rich environments. The high-frequency raw sensor data will be analyzed to obtain markers of tasks (e.g., typing), stress, activity, smoking, eating, sleep, alertness, and driving.

The researchers will employ state-of-the-art machine learning algorithms and novel model designs to overcome problems pervasive across sensor studies, such as occasional sensor data loss, model personalization, and lab-to-field generalization.

mPerf uses of a software platform developed by MD2K that collects, processes, and stores data. The mCerebrum smartphone suite includes 20+ apps that allows researchers to run multiple, concurrent field studies. A cloud component, Cerebral Cortex, analyzes data for biomarker discovery. To achieve these aims, mPerf has assembled a strong, multidisciplinary team representing the nation’s leading researchers in mobile sensing, machine learning, signal processing, behavior modeling, and human performance assessment.


This figure shows the process of converting high-frequency raw sensor data into markers that are used to generate indicators of job performance.