Equal Error Rate (EER)
Definition
The point on a classifier's detection error tradeoff curve where the false accept rate equals the false reject rate. Lower EER indicates better discrimination between genuine and spoofed speech. Used alongside t-DCF as a secondary metric in ASVspoof evaluations.
- Context
- Standard performance metric for biometric systems, especially speaker recognition and behavioral biometrics.
- What it measures
- The operating point where false acceptance rate equals false rejection rate.
- How to interpret it
- Lower EER indicates better discrimination and overall system performance.
Common questions
What does 'equal error rate' actually measure?+
EER is the operating point where a biometric system's false acceptance rate (incorrectly accepting an unauthorized person) equals its false rejection rate (incorrectly rejecting an authorized person). It's used to summarize system performance in a single number, so you can compare different biometric systems or configurations quickly.
Why does EER matter if false acceptances and rejections are different problems?+
In practice, you can adjust a biometric system's sensitivity up or down, trading off false acceptances against false rejections. EER shows the sweet spot where both error types are balanced. A lower EER means the system discriminates better overall, whether you're using speaker recognition or keystroke dynamics.
Is EER used outside speaker recognition?+
EER appears across all biometric systems that trade off false acceptances and rejections. It's used for speaker recognition, keystroke recognition, gait analysis, mouse dynamics, and other behavioral biometrics. NIST uses it as a standard metric for speaker recognition performance.
Related terms
- Anti-Spoofing Countermeasure (CM)
- A classifier, also called a CM system, trained to output a score indicating the probability that a given audio segment is genuine...
- ASVspoof
- A recurring evaluation campaign and dataset series that benchmarks anti-spoofing countermeasures against corpora of genuine and spoofed utterances. Editions in 2015, 2017,...
- Background Population Database
- A corpus of speaker recordings used to calibrate the LR in a forensic comparison. Must be matched to the case in language,...
- BATVOX
- A commercial automated speaker recognition system (originally Agnitio, later Nuance Communications) widely used in operational forensic casework in Europe and Latin America....
- C_llr (Log-Likelihood Ratio Cost)
- A proper scoring rule measuring the calibration quality of a speaker recognition system's LR outputs; a well-calibrated system has C_llr close to...
- Continuous Authentication
- An authentication approach that verifies user identity throughout an active session rather than only at login, using passively collected behavioural signals to...
- Covariate Problem
- In forensic gait analysis and computer-vision gait recognition, the degradation in accuracy caused by variations in viewpoint, clothing, footwear, walking speed, surface,...
- Dwell Time
- The period between an attacker gaining initial access and their detection. Reducing dwell time is a primary goal of threat hunting. The...
- ENFSI BPM
- The ENFSI Best Practice Manual for Forensic Comparison of Speech (2015, rev. 2022); the operational standard for speaker comparison in European forensic...
- ENFSI ENGAGE
- The European Network of Forensic Science Institutes' Gait Analysis across Europe project (2013-2015), which produced the principal best-practice manual for forensic gait...
- ENFSI Verbal Scale
- A standardised table published by the European Network of Forensic Science Institutes that maps ranges of LR values to verbal expressions of...
- Flight Time
- The interval between the key-up event of one key and the key-down event of the next key in a keystroke sequence. Also...
Explained in these topics
- Behavioural Biometrics: Keystroke, Gait and Mouse DynamicsThe operating point of a biometric system at which the false-acceptance rate equals the false-rejection rate. A standard single-number performance metric; lowe...
- Modern Automated Speaker Recognition: NIST SRE and the ENFSI BPMThe operating point where a speaker recognition system's false acceptance rate equals its false rejection rate. A standard NIST SRE performance metric; lower E...
- Voice Conversion and Cloning DetectionThe point on a classifier's detection error tradeoff curve where the false accept rate equals the false reject rate. Lower EER indicates better discrimination...