Source-Filter Vocoding and WORLD Quiz

Answers and explanations appear as you take the quiz.

1. In the source-filter model of speech, what does the 'filter' represent?
   • The resonant shape of the throat, mouth, and nasal cavity
   • The vibration rate of the vocal cords
   • The background noise in a recording
   • The microphone frequency response
2. Which three parameters does the WORLD vocoder extract from speech?
   • Volume, tempo, and reverb
   • F0 (pitch), spectral envelope, and aperiodicity
   • Phonemes, words, and pauses
   • Bitrate, sample rate, and channels
3. What is the name of WORLD's algorithm for estimating the spectral envelope?
   • WaveNet
   • MFCC
   • CheapTrick
   • Griffin-Lim
4. What does WORLD’s parameter separation allow an editor to do directly?
   • Convert speech to text without an ASR model
   • Remove the need for an audio input
   • Guarantee unchanged speaker identity after any pitch shift
   • Change the F0 stream while holding the spectral-envelope stream fixed
5. Which statement best describes WORLD compared with a neural vocoder?
   • WORLD is a signal-processing vocoder with explicit parameters; neural vocoders learn waveform generation
   • WORLD always sounds more natural than neural vocoders
   • WORLD is a language model that predicts phonemes
   • Both methods are identical parameter-free filters
6. What does the 'aperiodicity' parameter capture?
   • The loudness of the recording
   • The balance between tonal (periodic) and noisy (aperiodic) components
   • The number of speakers present
   • The language being spoken
7. Which WORLD component estimates bandwise aperiodicity?
   • CheapTrick
   • DIO
   • D4C
   • Harvest
8. Which WORLD method is an F0 estimator?
   • D4C
   • CheapTrick
   • Griffin-Lim
   • DIO