In this paper we present the application of sound environment analysis algorithms previously developed for ADL recognition on real recordings made in a medical nursing home. Initially the sound algorithms were tested in laboratory, secondly in a living lab in Grenoble and lastly in a nursing home (EHPAD). Several days of audio signal have been recorded, but for the moment only 24h are labelized and used for evaluation. The proposed system is based on a combination of Wavelets Transform, Gaussian Mixture Models (GMM) and Suport Vector Machines (SVM