Accelerometers enable us to analyse gait outside conventional gait laboratories. Before these devices can be used in large scale studies and in clinical settings a thorough evaluation of their performance in different populations is required. The aim of this study was to present an acceleration-based reference database for healthy gait. The repeatability and inter-observer reliability of acceleration-based gait analysis was investigated. The sensitivity was tested on different age groups and the effect of gender was studied. A comprehensive set of gait parameters (i.e. cadence, speed, asymmetry and irregularity) were studied in 60 women and 60 men. Basic gait parameters showed high repeatability (VC(cadence) 1.51%, ICC(cadence) 0.996) and inter-observer reliability (ICC(cadence) 0.916), while asymmetry and irregularity showed lower repeatability (VC(asym) 47.88%, ICC(asym) 0.787) and inter-observer reliability (ICC(asym) 0.449). The effects of age and gender on gait parameters were found to be consistent with those reported in studies using other methodologies. These findings and the advantages of the device support the application of AGA for routine clinical use and in daily life.