You are on page 1of 5

과제 LAB 13 :

Noise Reduction
Speech Enhancement (1)
Digitally recorded speech in a noisy environment
can be enhanced using a lowpass filter if the
recorded speech with a sampling rate of 8,000 Hz
contains the desired frequency components lower
than 1,500 Hz.
Design a lowpass filter to remove the high
frequency noise above 1,500 Hz with the following
filter specifications:
– Passband frequency: 0 – 1,500 Hz
– Passband ripple: 0.1 dB (3 dB)
– Stopband frequency range: 2,000 – 4,000 Hz
– Stopband attenuation: 50 dB (20 dB)

LAB 13 과제: Noise Reduction 2


Speech Enhancement (2)
Use the designed lowpass filter to filter the noisy
speech and adopt the following code to simulate the
noisy speech:

>> load speech.dat


>> t=[0:length(speech)]*T;
>> noise_power=mean(speech.*speech)/4; % Noise power
=(1/4) speech power
>> noise=sqrt(noise_power)*randn([1,length(speech)]); %
Generate Gaussian Noise
>> nspeech=speech+noise; % Generate noisy speech

LAB 13 과제: Noise Reduction 3


Speech Enhancement (3)
1. Plot the speech samples and one-sided amplitude
spectra for the original speech and the noisy
speech.
2. Design the following digital filters, print the filter
coefficients, plot the frequency response for each
filter, and verify the results.
– Passband ripple: 0.1 dB, Stopband attenuation: 50 dB
– Passband ripple: 3 dB, Stopband attenuation: 20 dB
a. Optimal filter using firpmord
b. Butterworth digital filter using butter and buttord

LAB 13 과제: Noise Reduction 4


Speech Enhancement (4)
3. Analyze the system performance
a. Objective Testing (SNR)
b. Subjective Testing: Evaluate the filter performance by
listening to the filtered signals.
4. Discuss about the results with respect to two
different specifications and two different filters.

LAB 13 과제: Noise Reduction 5

You might also like