PLDA (Probabilistic Linear Discriminant Analysis)
Definition
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 variability and between-speaker variability in the training population.
- Model type
- Statistical backend for speaker comparison
- Input
- Speaker embeddings (from i-vector or x-vector systems)
- Output
- Log-likelihood ratio for same-speaker vs. different-speaker hypotheses
Common questions
What does PLDA do in speaker comparison?+
PLDA (Probabilistic Linear Discriminant Analysis) is a statistical model that takes two speaker voice samples and scores how similar they are. It computes a log-likelihood ratio between the hypothesis that both samples came from the same person versus different people, accounting for natural variation within a single speaker's voice and differences between speakers.
How does PLDA handle speaker variability?+
PLDA explicitly models two types of variability: within-speaker variability (how much a single person's voice changes across different recordings) and between-speaker variability (how speakers differ from each other). This normalization helps the model give fair similarity scores even when speakers have unusually consistent or highly variable voices.
Which voice identification systems use PLDA?+
PLDA works as a backend model paired with speaker embedding systems like i-vectors and x-vectors. These embeddings compress voice characteristics into numerical representations, which PLDA then compares to generate a similarity score suitable for forensic speaker identification work.
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...
- Forensic Speaker Comparison
- A systematic examination comparing acoustic and phonetic features of a questioned voice recording against known reference recordings of a named individual, expressed...
- Formant Frequencies
- Resonance frequencies of the vocal tract that shape vowel quality. F1 and F2 (the first and second formants) are the most informative...
- 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....
- I-Vector / X-Vector
- Fixed-length mathematical representations of a speech utterance used in automatic speaker recognition. I-vectors are derived from Gaussian mixture model statistics; x-vectors are...
- IAFPA
- International Association for Forensic Phonetics and Acoustics: the primary professional body for forensic phoneticians and audio analysts, which publishes guidelines on speaker...
- Likelihood Ratio (LR)
- The ratio of two conditional probabilities: the probability of the observed evidence given the prosecution's hypothesis (same source), divided by the probability...
Explained in these topics
- Modern Automated Speaker Recognition: NIST SRE and the ENFSI BPMA backend model for speaker comparison that explicitly models between-speaker and within-speaker covariance, producing a log-likelihood ratio between the same-...
- Speaker Comparison: Methods and the Expert's RoleA statistical back-end model used with i-vector and x-vector systems to compute a similarity score between two utterance representations, normalised for within...