Falls are a well-known danger for older adults. With the worldwide population aging, there has been an increasing interest in assessing the risk of falling. This work presents a novel algorithm for continuous fall risk assessment, relying on a linear regression model whose inputs consist of both measured and self-reported risk factors. Two models were conceived and compared, following two distinct approaches, a theoretical and an empirical one. The system is pervasive and was tested in free-living unsupervised conditions. The results of our fall risk scoring system unveiled a strong correlation with the output of the clinical functional tests POMA and TUG (90% and 89%, respectively), which was deemed a promising outcome concerning the feasibility of pervasive monitoring for fall risk assessment in daily living.