Writer identification and verification have gained increased interest recently, especially in the fields of forensic document examination and biometrics. Writer identification assigns a handwriting to one writer out of a set of writers. It determines whether or not a given handwritten text has in fact been written by a claimed writer. While in the offline case, the data contains spatial information about the handwriting only, in the online case, the data additionally encodes temporal information. The following contributions are made in this work: in the offline case, two state-of-the-art systems to address the tasks of writer identification and verification are developed. Different confidence measures are defined to assess the quality of the recognition. The performances of the two systems are evaluated and compared on a large data set. Furthermore, feature selection methods are applied to improve the performance of an existing writer identification system. In the online case, a system to identify the writer of online hand-written notes on a whiteboard is developed. Different feature sets are defined and a new approach to improve the performance of a writer identification system by fusing the asynchronous feature streams is presented. Various parameters influencing the performance of the system are systematically studied on a large data set.