An integrated workstation developed for use in clinical medicine is described. The specific application domain is gastroenterology with endoscopic, ultrasonic, and laboratory units. The following modules for data collection and management during ultrasound examinations have been implemented and tested: a multimodal user interface with speech and symbol recognition to reduce complexity in human-machine interaction, knowledge-based components to support the user and to structure dialogues dynamically, and techniques for handling multimedia documents that consist of images, text, structured text, graphics, and digitized speech. The system was implemented using networked frontend and backend workstations under UNIX.>

Recently there has been increasing attention to character recognition/graphical user interfaces under the name of “gesture input”. This technique actually has a long history: “sketch recognition” interfaces of 15 or more years ago were highly praised [Applicon 73], and user interfaces using handwriting input before the wide use of text keyboards were one of the first research goals in computer science [Bledsoe 59]. The underlying character and symbol recognition technologies have been a major research area in their own right since the early 1950s [Suen 80].

The paper describes the design and the optimization of a contextual postprocessing system (CPPS) operating in combination with a multiple-choice single-character recognition system (SCRS). The CPPS is part of the automatic postal-address reading machine for the Federal German Postal Service.