In this paper we introduce a novel device-free radio based activity recognition with localization method with various applications, such as e-Healthcare and security. Our method uses the properties of the signal subspace, which are estimated using signal eigenvectors of the covariance matrix obtained from an antenna array (array sensor) at the receiver side. To classify human activities (e.g., standing and moving) and/or positions, we apply a machine learning method with support vector machines (SVM). We compare the classification accuracy of the proposed method with signal subspace features and received signal strength (RSS). We analyze the impact of antenna deployment on classification accuracy in non-line-of-sight (NLOS) environments to prove the effectiveness of the proposed method. In addition, we compare our classification method with k-Nearest Neighbor (KNN). The experimental results show that the proposed method with signal subspace features provides accuracy improvements over the RSS-based method.