Speaker Recognition & Verification
ASR asks "what did they say?" Speaker recognition asks "who said it?" The math looks the same — embeddings plus cosine — but every production decision hinges on a single EER number. A user says a passphrase. You want to know: is this the person they claim to be (verification, 1:1), or is it the first person in your enrollment bank (identification, 1:N)? Or neither — is this an unknown speaker (open-set)? Pre-2018: GMM-UBM + i-vectors. Reasonable EER but fragile to channel shift (phone vs laptop) and emotion. 2018–2022: x-vectors (TDNN backbone trained with angular margin). 2022+: ECAPA-TDNN and WavLM-large embeddings. By 2026 the field is dominated by three models and one metric. The metric is EER — Equal Error Rate. Set your decision threshold so False Accept Rate = False Reject Rate. The crossover is EER. Used in every paper, every leaderboard, every procurement call. Enrollment + verification pipeline with embedding + cosine + EER The pipeline. Enrollment: record 5–30 seconds of the target speaker; compute a fixed-dimension embedding (192-d for ECAPA-TDNN, 256-d for WavLM-large). Verification: get the test utterance embedding; compute cosine similarity; compare to a threshold. ECAPA-TDNN (2020, still dominant 2026). Emphasized Channel Attention, Propagation and Aggregation - Time-Delay Neural Network. 1D conv blocks with squeeze-excitation, multi-head attention pooling, followed by a linear layer to 192-d. Trained…
Speaker Recognition & Verification: ASR asks "what did they say?" Speaker recognition asks "who said it?" The math looks the same — embeddings plus cosine —…
This free lesson is part of the AI Engineering from Scratch curriculum. Read the full explanation, run the lesson code, and verify the result in the interactive reader or from the repository source.
Browse the complete course catalog or open this lesson on GitHub.