Scribd
Upload
44 views5 pages

Sample Questions UCS749

The document contains a series of questions related to signal processing, including topics such as sampling rates, Nyquist theorem, quantization levels, Short-Time Fourier Transform, Hidden Markov Models, and speech recognition techniques. It also addresses statistical estimation methods like Maximum Likelihood Estimation for analyzing data samples. Additionally, it discusses the relevance of the Mel scale and delta coefficients in feature extraction for speech signals.

Uploaded by

akshita03sharma
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
44 views5 pages

Sample Questions UCS749

The document contains a series of questions related to signal processing, including topics such as sampling rates, Nyquist theorem, quantization levels, Short-Time Fourier Transform, Hidden Markov Models, and speech recognition techniques. It also addresses statistical estimation methods like Maximum Likelihood Estimation for analyzing data samples. Additionally, it discusses the relevance of the Mel scale and delta coefficients in feature extraction for speech signals.

Uploaded by

akshita03sharma
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 5

Q1.

A signal is sampled at 20 kHz, and you receive the following samples:

What are the actual frequency components present in the original signal?

Q2. An ECG (electrocardiogram) signal has a maximum frequency of 150 Hz. What is the
Nyquist rate and the minimum sampling interval required for accurate digital recording?

Q3. Calculate the total number of quantization levels if an 8 bit quantizer is used to convert
analog to digital signal.

Q4. A speech signal is analyzed using a Short-Time Fourier Transform (STFT) with a 50 ms
window. If we reduce the window length to 25 ms, what happens to the frequency
resolution?

Q5. A speech signal is sampled at 16 kHz and contains frequencies up to 7 kHz. To


compress the signal, we want to downsample it to 8 kHz. What should be the cutoff
frequency?

Q6. Consider an HMM with two states S1 and S2 and the following parameters:

What is the probability of the observation sequence O=(O1,O2,O1)?

Q7. What is the most likely sequence of hidden states using the Viterbi Algorithm for above
parameters?
Q8. An HMM has

After one iteration of the Baum-Welch Algorithm, what are the updated transition and
observation probabilities?

Q9. In a speech recognition system, a phoneme is modeled as an HMM with 5 states. A


speech signal is observed with a sequence length of 10. If the initial state probabilities are
uniform and the transition matrix follows a left-to-right model (each state can only transition
to itself or the next state), compute the probability of a valid path for the observation
sequence.

Q10. A quality control engineer is analyzing the weight of packets of rice produced by a
machine. The weights (in kg) of a sample of 10 packets are:
2.01,2.03,1.98,2.05,1.99,2.02,2.00,2.04,1.97,2.01
Estimate the Mean (μ) and Variance (σ2) using Maximum Likelihood Estimation (MLE).

Q11. Why is the Mel scale used in MFCC extraction?

The Mel scale is used because the human ear perceives frequencies non-linearly. Low
frequencies are perceived more distinctly than high frequencies. The Mel scale warps the
frequency axis to match human auditory perception, improving speech recognition
accuracy.

Q12. How does MFCC differ from PLP?

Q13. Why are delta coefficients used in MFCC?

MFCC captures spectral information, but speech sounds change over time. Delta
coefficients capture the rate of change of MFCCs.

Q14. Why is pre-emphasis applied before extracting MFCCs?

Pre-emphasis is a high-pass filter that boosts high frequencies, compensating for the
natural roll-off in the speech spectrum.

Q15. Compute the number of samples in a 25 ms frame at a 16 kHz sampling rate.


Q16. Compute Hamming window values for N=10

Q17. Explain the importance of equal-loudness pre-emphasis in PLP.

You might also like