X-Vector
Definition
A deep-learning speaker embedding produced by the penultimate layer of a time-delay neural network (TDNN) trained for speaker classification. Introduced by Snyder et al. (2018); dominant in NIST SRE from 2018 onwards due to superior cross-channel and cross-language generalisation.
- Architecture
- Time-delay neural network (TDNN)
- Extraction point
- Penultimate hidden layer embedding
- Introduced by
- Snyder et al., 2018
- Adoption
- Dominant in NIST SRE from 2018 onward
- Strength
- Better cross-channel and cross-language generalisation than i-vectors
Common questions
What did x-vectors replace in automated speaker recognition?+
X-vectors superseded i-vectors, the earlier statistical embedding built from a Gaussian mixture model universal background model, moving the field from a factor-analysis approach to a deep neural embedding trained discriminatively for speaker classification.
Why do x-vectors generalise better across channels and languages?+
The TDNN is trained on large, diverse speaker datasets with augmentation such as added noise and reverberation, forcing the embedding to capture speaker-specific vocal characteristics rather than channel or language artefacts, unlike i-vectors which were more sensitive to train-test mismatch.
Are x-vectors still the current standard?+
X-vectors set the standard from 2018 but have since been challenged by newer embeddings such as ECAPA-TDNN built on the same TDNN family, so current NIST SRE systems often use an evolved variant rather than the original x-vector design.
Related terms
- 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...
- 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 Verbal Scale
- A standardised table published by the European Network of Forensic Science Institutes that maps ranges of LR values to verbal expressions of...
- Equal Error Rate (EER)
- The point on a classifier's detection error tradeoff curve where the false accept rate equals the false reject rate. Lower EER indicates...
- I-Vector
- A fixed-length low-dimensional speaker representation extracted by projecting a speaker's UBM statistics into a total variability space. Introduced by Dehak et al....
- NIST SRE (Speaker Recognition Evaluation)
- A competitive benchmarking series run by the National Institute of Standards and Technology since 1996, in which research and commercial teams submit...
- PLDA (Probabilistic Linear Discriminant Analysis)
- A statistical back-end model used with i-vector and x-vector systems to compute a similarity score between two utterance representations, normalised for within-speaker...