Pitch (in speech) MATLAB tutorial series (Part 2.2) Pouyan - PowerPoint PPT Presentation

Pitch (in speech) MATLAB tutorial series (Part 2.2) Pouyan Ebrahimbabaie Laboratory for Signal and Image Exploitation (INTELSIG) Dept. of Electrical Engineering and Computer Science University of Liège Liège, Belgium Applied digital signal processing (ELEN0071-1) 27 Mars 2019

Speech production mechanism

Speech production mechanism

Speech production mechanism

Speech production mechanism Pitch of a person’s voice reflects the fundamental frequency which the vocal folds are vibrating.

“Pitch” in psychology of voice • Males voice pitch: 85 to 180 Hz • Females voice pitch: 165 to 255 Hz • Vocal folds are ticker and larger in males (A dam’s apple in males) • Larynx size partly controlled by testosterone • Voice pitch may dramatically influence our judgment about the speakers • Male with low-pitched voice tend to be perceived as more, • Attractive • Physically stronger • Dominant (respected & commanding) • Aggressive (very low-pitched)

“Pitch” in psychology of voice • Female with high-pitched voice tend to be perceived as more, • Attractive • Female with low-pitched voice tend to be perceived as more, • Dominant • Researches on many elections showed that females and males preferred low-pitched candidates

“Pitch” in psychology of voice • Female with high-pitched voice tend to be perceived as more, • Attractive • Female with low-pitched voice tend to be perceived as more, • Dominant • Researches on many elections showed that females and males preferred low-pitched candidates Simple test: which one you will vote for?

“Pitch” in psychology of voice • Female with high-pitched voice tend to be perceived as more, • Attractive • Female with low-pitched voice tend to be perceived as more, • Dominant • Researches on many elections showed that females and males preferred low-pitched candidates Simple test: which one you will vote for?

Application: speech recognition

Application: speech recognition

MATLAB: Pitch MATLAB R2018b: [f0,loc]=pitch(audioIn,fs)

MATLAB: Pitch MATLAB R2018b: [f0,loc]=pitch(audioIn,fs) Methods: NCF – Normalized Correlation Function PEF – Pitch Estimation Filter …

MATLAB: Pitch

Example 2.6: Pitch 1 % Read audio file [y,fs] = audioread('TNS_Pitch1.mp3'); % Play the sound sound(y,fs) % Time vector t=0:1/fs:(length(y)-1)/fs; % Plot the sound figure(1) plot(t,y,'LineWidth',1) ylabel('Amplitude') xlabel('Time (second)')

Example 2.6: Pitch 1 % Extract pitces and their corresponding indexes [f0,idx] = pitch(y,fs); % Plot pitches figure(2) plot(idx,f0,'LineWidth',2) ylabel('Pitch (Hz)') xlabel('Sample Number')

Example 2.6: Pitch 1 % Find trend (using Moving Average fillter) n=700; wts = [1/n;repmat(1/n,n-1,1);1/n]; f0S = conv(f0,wts,'valid'); % Plot trend figure(3) plot(f0S,'LineWidth',2)

Example 2.7: Pitch 2 % Read audio file [y] = audioread('TNS_Pitch2_A.mp3'); % Pick a channel yA=y(:,1); % Read audio file [y,fs] = audioread('TNS_Pitch2_B.mp3'); % Pick a channel yB=y(:,1); % Pitch A [f0A,idxA] = pitch(yA,fs); % Pitch B [f0B,idxB] = pitch(yB,fs);

Example 2.7: Pitch 2 % Find trends (using Moving Average fillter) n=700; wts = [1/n;repmat(1/n,n-1,1);1/n]; f0SA = conv(f0A,wts,'valid'); f0SB = conv(f0B,wts,'valid'); % Plot trend figure(1) plot(f0SA,'LineWidth',2) Pause() … plot(f0SB,'LineWidth',2) …

It is popular among some politicians, actors and singers to lower their pitch ~ - 40 Hz

Useful links • https://nl.mathworks.com/help/audio/ref/pitch.html • https://nl.mathworks.com/help/econ/moving-average- trend-estimation.html • https://www.researchgate.net/publication/314580952_ Psychology_of_Voice • https://nl.mathworks.com/help/econ/moving-average- trend-estimation.html • https://www.americanscientist.org/article/how-voice- pitch-influences-our-choice-of-leaders • https://www.ncbi.nlm.nih.gov/pmc/articles/PMC335071 3/