Design Automation of High Speed Digital Signal Processing in VLSI with Applications in Speech Recognition Systems Based on Hidden Markov Models

The rapidly growing complexity and increasing performance of integrated circuit technology has inspired the work on high performance CMOS circuit design, presented in this thesis. Predicting the future of IC technology, the SIA Roadmap, which was presented two years ago by the Silicon Industry Association, is presently proving to be on the conservative side. Dealing with large and complex designs, and at the same time designing circuits for high performance applications is a major challenge. As an example of demanding digital signal processing, the area of automatic speech recognition is utilized as a driving application throughout the text of this dissertation. Speaker independent speech recognition tasks with large vocabulary are at focus, and the well established statistical methods of Hidden Markov Models are utilized. In this work, design methodologies as well as cell libraries for utilization of the dynamic single phase clocking techniques TSPC (True Single Phase Clocking circuit..