O (10)

Question 1

What is the first step in the short-term spectrum process?

Answer 1

apply Hamming window

This step prepares the signal for Fourier transform analysis.

Question 2

What is computed after applying the Hamming window in the short-term spectrum?

Answer 2

compute Fourier transform (as an FFT)

FFT stands for Fast Fourier Transform, which is an efficient algorithm for computing the Fourier transform.

Question 3

What is taken from the coefficients after computing the Fourier transform in the short-term spectrum?

Answer 3

take magnitude of coefficients

This step extracts the amplitude information from the frequency components.

Question 4

What are the two types of binning mentioned?

Answer 4

• Mel (pitch perception) & triangular
• Bark (critical band) & trapezoidal

Binning helps in grouping frequency components for analysis.

Question 5

What does frequency weighting include?

Answer 5

• A zero giving +6 dB/octave pre-emphasis
• Equal-loudness weighting

Frequency weighting adjusts the amplitude of frequencies to match human perception.

Question 6

What are the two methods of compression listed?

Answer 6

• logarithm
• cube root (intensity-loudness power law)

Compression techniques are used to reduce the dynamic range of audio signals.

Question 7

What is the purpose of the inverse FT?

Answer 7

information concentration

This process reconstructs the signal from its frequency components.

Question 8

What are the two methods used in the inverse FT?

Answer 8

• IDCT
• IDFT

IDCT stands for Inverse Discrete Cosine Transform, and IDFT stands for Inverse Discrete Fourier Transform.

Question 9

What does spectral smoothing involve?

Answer 9

• cepstral truncation
• AR model (i.e., LP)

Spectral smoothing reduces fluctuations in the spectral representation of signals.

Question 10

What is the status of decorrelation in the process?

Answer 10

• none required
• conversion to cepstra

Decorrelation is sometimes necessary to simplify the feature representation.

Question 11

What is the purpose of liftering?

Answer 11

variance equalisation

Liftering adjusts the variance of cepstral coefficients to improve feature representation.

Question 12

What are the two types of features compared?

Answer 12

• MFCC
• PLP-CC

MFCC stands for Mel-Frequency Cepstral Coefficients, and PLP-CC stands for Perceptual Linear Prediction Coefficients.

Question 13

What is the operation used in STFT?

Answer 13

Frequency

STFT stands for Short-Time Fourier Transform, which analyzes the frequency content of signals over time.