Although by training I am an electrical engineer, my PhD dissertation is in bio-stats (more specifically multiple testing). Therefore, I have been forced to do some thinking about the issues of detection and estimation both from performance in engineering systems and from scientific inference point of view. By now, I have done statistical work in bunch of different fields -wireless comm, speech recognition, machine learning, bioinformatics, and biostatistics. While I learned dectection and estimation for the first time, which was 5-6 years ago, I did not use Kay's book, but when couple of years ago I was beating my head over certain Frequentist/Bayesian differences, I accidentally ran into this book that lies in my Prof's collection.
Kay is plain good. He seem to have an amazingly clarity about these issues. His treatment is sometimes too direct and simple that you feel there must be some catch. But now I am increasingly getting convinced that there is none. This book surmounts much confusion that is ingrained to statistical literature of this level and scope. I have read a lot of references but I don't know many that make Neyman-Pearson Lemma or Cramer-Rao bound so clear and straightforward. Summaries at the beginning of each chapter (yes beginning) are right-on-the-button especially for those who have been in the field for a while and still occasionally need coordinates. The book is perfect as a textbook in a comprehensive graduate course on estimation-detection or as a handy reference.
* If download links doesn't work. Please write a comment.Fundamentals Of Statistical Signal Processing (2 Volumes) Download via usenet