Speaker Verification Quiz

Answers and explanations appear as you take the quiz.

1. Speaker verification is best described as which type of decision?
   • One-to-one: does this voice match the claimed identity?
   • One-to-many: whose voice is this?
   • Transcribing speech to text
   • Detecting silence in audio
2. What is the difference between text-dependent and text-independent verification?
   • One uses video, the other audio
   • Text-dependent requires a fixed passphrase; text-independent works on any speech
   • Text-dependent is for music only
   • There is no difference
3. What does the Equal Error Rate (EER) represent?
   • The total number of users
   • The audio sample rate
   • The point where false accepts equal false rejects
   • The length of the passphrase
4. Why is anti-spoofing important in speaker verification?
   • To compress audio faster
   • To translate languages
   • To increase passphrase length
   • To defend against recordings and AI-generated deepfake voices
5. What is a stored "voiceprint" used for in verification?
   • Serving as the enrolled reference the test voice is compared against
   • Playing back the user's speech
   • Compressing the audio file
   • Translating speech to text
6. Lowering the acceptance threshold typically does what?
   • Rejects more genuine users and blocks more impostors
   • Accepts more impostors while accepting more genuine users
   • Has no effect on errors
   • Only changes the audio length
7. What does an ASVspoof evaluation test in addition to ordinary speaker matching?
   • Whether the speaker’s words are grammatically correct
   • Whether the microphone has a camera
   • Whether countermeasures can detect replayed or synthesized spoofed speech
   • Whether the voice belongs to a known language
8. Why does a speaker match score not prove that a live person spoke the audio?
   • Speaker embeddings contain the speaker’s legal name
   • A match score measures camera presence
   • Verification does not analyze audio
   • A recording or synthetic voice may imitate the enrolled speaker