Spectral Subtraction and Wiener Filtering Quiz

Answers and explanations appear as you take the quiz.

1. What annoying artifact is commonly associated with spectral subtraction?
   • Musical noise (random fleeting tones)
   • Echo cancellation
   • Volume normalization
   • Pitch shifting
2. In which domain do spectral subtraction and Wiener filtering primarily operate?
   • The raw time domain only
   • The frequency domain after an STFT
   • The text domain
   • The image domain
3. Under a specified statistical signal-and-noise model, what criterion does the classical Wiener estimator minimize?
   • The number of STFT frames
   • The file’s bit rate
   • Expected mean-squared estimation error
   • The number of microphones
4. How does spectral subtraction typically estimate the noise spectrum?
   • From a random number generator
   • From the text transcript
   • From the loudest part of the signal
   • From silent or speech-free segments of the audio
5. In the common power-spectrum Wiener gain G=Ps/(Ps+Pn), what do Ps and Pn denote?
   • Estimated clean-signal and noise powers in that frequency bin
   • Speech and noise sample counts
   • Two different phase angles
   • The number of filters and frames
6. What key assumption limits both methods against sudden sounds like a door slam?
   • That speech is always loud
   • That the noise is relatively stationary
   • That there is only one speaker
   • That the audio is in stereo
7. What may happen to the estimated magnitude in a spectral-subtraction bin when the noise estimate exceeds the observation?
   • It is wrapped into a positive phase
   • It is replaced by a transcript token
   • The negative result is floored to zero
   • It is doubled to preserve loudness
8. When a real-valued Wiener gain is applied to a complex noisy STFT coefficient, what is commonly done with its phase?
   • Phase is replaced by the text prompt
   • Phase is always set to zero
   • The coefficient is converted into a stereo channel
   • The noisy phase is retained while the magnitude is scaled