Small variation in amplitude

Does anybody can tell me plz,how can I make change in the amplitude of speech signal ?I need to make small variation in the amplitude.I am working in synthesize speech ,I would like to change the amplitude of the speech signal signal.

Answer 1

[Wayne King]

You can amplitude modulate the different sections.

For example, I'll amplitude modulate from 1 to 3 seconds in the following signal

Fs = 1000;
t = 0:1/Fs:5-(1/Fs);
x = cos(2*pi*100*t);
x(1000:3000) = x(1000:3000).*(1+0.2*cos(2*pi*10*t(1:2001)));

Vary the modulation depth (I've used 0.2 here) and the modulation frequency (here 10 Hz) to get a realistic variation in amplitude.

The modulation depth should be between 0 and 1, I'm guessing you want a small value like 0.1, 0.2.

The modulation frequency should probably be a low frequency as well maybe 5 Hz or less.

[Farideh Jalali]

This is very good idea but the major problem is that I cant consider the the Cos model as excitation for my speech synthesize model.I tried to modified the codes as following but still I cant hear any variation in the amplitude of voice signal.

fs=10000;    % sampling frequency (Hz)
NS = 3*fs;  % Number of samples to be generated (for 3 seconds)
 pp=100
 fm = 10; 
 fc = fs/pp ; 

for m=1:NS,
 y(m)=cos(2*pi*(fc/fs)*m).*(1+0.2*cos(2*pi*(fm/fs)*m));

end;
figure (1); plot (y(1:3600));

[Pks,Locs] = findpeaks (y, 'minpeakheight', 0.7);
impEx = zeros (1,NS);
for m=1:length (Locs), impEx (Locs (m)) = 1; end;
figure (66),plot (impEx(1:800) );

Answer 2

[Wayne King]

Just multiply the signal by a constant between 0 and 1

    t = linspace(0,1,1000);
    x = cos(2*pi*100*t)+sin(2*pi*100*(t-pi/4));
    % scale the amplitude by 1/5
    y = 0.2*x;
    plot(t,x), hold on;
    plot(t,y,'r','linewidth',2)

If you want to add just a small variation, then multiply by a factor just slightly over or under 1.

[Farideh Jalali]

Thank you very much indeed.When I am listening to the voice,I should hear variation in different time of voice.With your method I only can see changes in the amplitude of the whole voice signal.For example,something ,I expect is :when you listen to the patient voice which is trying to say /a/ in 3 second.you can hear amplitude variation in each second of her voice.I have to make these small variations in my synthesize speech.In each part of voice these variation should be small and can be different from other section of voice.Do you have any suggestion for Am demodulation?

Answer 3

[Farideh Jalali]

This is very good idea but the major problem is that I cant consider the the Cos model as excitation for my speech synthesize model.I tried to modified the codes as following but still I cant hear any variation in the amplitude of voice signal.

fs=10000; % sampling frequency (Hz) NS = 3*fs; % Number of samples to be generated (for 3 seconds) pp=100 fm = 10; fc = fs/pp ;

for m=1:NS, y(m)=cos(2*pi*(fc/fs)*m).*(1+0.2*cos(2*pi*(fm/fs)*m));

end; figure (1); plot (y(1:3600));

[Pks,Locs] = findpeaks (y, 'minpeakheight', 0.7); impEx = zeros (1,NS); for m=1:length (Locs), impEx (Locs (m)) = 1; end; figure (66),plot (impEx(1:800) );