Signal Quantization and Compression Exercises
-
Select 5 bits and the male sound. Play both the original and the quantized
sounds when truncation, rounding, and sign magnitude truncation are used.
Which quantization method leads to the best quality sound after quantization?
-
Select 12 bits. For each sound signal (female, male), detemine
the resulting SNR when rounding, truncation, or sign magnitude truncation
is used. Collect the values into a table. Verify that the method of rounding
leads to the highest SNR? For the remainder of these exercises, we
shall use rounding only.
-
For every sound signal (female or male), what is the minimum number of bits
that you feel is required to have a good quality sound for quantized signal?
What are then the corresponding bit rates assuming that the signals are
sampled at 8000 samples per second? What are the SNR in each case for the
minimum number of bits that you chose? Are they all equal?
-
For both male and female speech signals, complete a table
with the SNR values that are obtained for quantization with 5,7,9,12 bits.
Plot the numbers on the same graph, with one curve for the male and another
for the female speaker. What do you notice? What improvement in terms of
SNR can we expect from adding one bit in the quantizer?
-
Consider again the male and the female speech signals and encode them with
5 bits. Notice that the quality is not good enough for a telephone conversation.
Apply the mu-law companding technique and find for both signals the value
of mu (from the ones available in the experiment) that maximizes the SNR. Is the quality of the new quantized signals
acceptable?
-
Consider again the female speech signal and quantize it with 2 bits without
companding. Can you see the quantization levels on the plot?
Explain why we hear some silences in the quantized signal.
-
Choose 5 bits with truncation and listen to the quantized signal. Now,
choose rounding and mu companding with mu=100. Listen again to the quantized
signal and notice the difference.