Elementary Signal Detection Theory
Signal detection theory, as developed in electrical engineering and based on statistical decision theory, was first applied to human sensory discrimination about 40 years ago. The theory's intent was to explain how humans discriminate and how we might use reliable measures to quantify this ability. An interesting finding of this work is that decisions are involved even in the simplest of discrimination tasks--say, determining whether or not a sound has been heard (a yes-no decision). Detection theory has been applied to a host of varied problems (for example, measuring the accuracy of diagnostic systems, survey research, reliability of lie detection tests) and extends far beyond the detection of signals. This book is a primer on signal detection theory, useful for both undergraduates and graduate students.
Why Read This Book
You should read this book if you want a clear, accessible introduction to the concepts behind detection performance and ROC analysis without getting lost in heavy math. It gives you the intuition and practical measures (d', criterion, ROC) you need to evaluate and compare detectors and experimental results.
Who Will Benefit
Graduate students and practicing signal-processing engineers who need a compact, concept-driven introduction to statistical detection and ROC-based performance evaluation.
Level: Intermediate — Prerequisites: Basic probability and statistics (mean, variance, Gaussian distribution), familiarity with hypothesis testing and elementary calculus.
Key Takeaways
- Understand and interpret ROC curves and area-under-curve as detector performance metrics.
- Compute and use sensitivity measures such as d' and criterion measures for bias.
- Apply the likelihood-ratio and Neyman–Pearson framework to binary detection problems.
- Estimate detection-performance parameters from experimental/empirical data.
- Differentiate between yes–no, forced-choice, and rating-scale procedures and their effects on measured sensitivity.
- Relate psychophysical detection measures to engineering detection metrics used in communications and radar.
Topics Covered
- Introduction to Signal Detection Theory
- Basic Concepts: Noise, Signal+Noise, and Decision Rules
- ROC Curves: Construction and Interpretation
- Measures of Sensitivity and Bias (d', β, A')
- Yes–No, Forced-Choice, and Rating Procedures
- Gaussian SDT Models: Equal- and Unequal-Variance Cases
- Parameter Estimation and ROC Smoothing
- Nonparametric and Robust Measures
- Applications and Experimental Design Considerations
- Extensions: Multi-interval and Multidimensional Detection
- Practical Examples and Data Analysis
How It Compares
Covers much the same conceptual ground as Macmillan & Creelman's Detection Theory but is more concise and pedagogical; for a more mathematical/engineering treatment of detection and estimation, see Steven M. Kay's Fundamentals of Statistical Signal Processing, Volume 2 (Detection Theory).
