Real-time detection of freezing of gait in Parkinson’s disease using multi-head convolutional neural networks and a single inertial sensor

•Freezing of gait (FOG) is a disabling symptom of Parkinson’s disease (PD).•Accurate real-time detection of FOG during daily activities is a challenging task.•Data were recorded from 118 PD and 21 elderly subjects using a single accelerometer.•The developed multi-resolution CNN can accurately detect and even predict FOG.•The algorithm is robust, fast, and lightweight.