Pitch is the fundamental frequency of vibration of the vocal folds, which are present at the top of one's trachea. They vibrate quasi-periodically only for voiced phonemes, namely vowel, semivowel and nasal sounds. So, for unvoiced stops such as /p/, /k/, /t/, /th/, /ch/ and unvoiced fricatives such as /f/, /s/, etc. there is nothing called pitch.
The formant frequencies are due to the frequency shaping of the signal from the vocal folds by the vocal tract. Vocal tract is everything from nasal tract, tongue, teeth, lips, palate, etc. The particular configuration of the above organs (articulators) for every phoneme creates resonances at specific frequencies called formants. So, formants exist for both voiced and unvoiced sounds.
Pitch can be estimated by quantifying the period (using autocorrelation, say) or measuring the harmonics. Formant frequecies can be found by linear prediction analysis from the poles.
Pitch is the fundamental frequency of vibration of the vocal folds, which are present at the top of one's trachea. They vibrate quasi-periodically only for voiced phonemes, namely vowel, semivowel and nasal sounds. So, for unvoiced stops such as /p/, /k/, /t/, /th/, /ch/ and unvoiced fricatives such as /f/, /s/, etc. there is nothing called pitch.
The formant frequencies are due to the frequency shaping of the signal from the vocal folds by the vocal tract. Vocal tract is everything from nasal tract, tongue, teeth, lips, palate, etc. The particular configuration of the above organs (articulators) for every phoneme creates resonances at specific frequencies called formants. So, formants exist for both voiced and unvoiced sounds.
Pitch can be estimated by quantifying the period (using autocorrelation, say) or measuring the harmonics. Formant frequecies can be found by linear prediction analysis from the poles.
Thank you very much sir ...
Sir may i know what are the various factors that cause variation in formant frequency ?
eg.. age , sex(m/f) ... are there any others ? can bmi or height or weight or pitch or emotion be any of those factors ?
Dear Venkatakrishnan Muralidaran, Pitch is the fundamental frequency (steady) whereas, Formants are strong frequency peaks (transients).
Formant frequencies mainly vary due to the phoneme being articulated and not so much by factors you mentioned such as age, weight, etc.
Can we say that pitch is due to periodicity and formant due to resonance ?
Pitch and Frequency are also influenced by the affective state of the user, see e.g.
Bogdan Vlasenko, Dmytro Prylipko, Ronald Böck, Andreas Wendemuth
Modeling phonetic pattern variability in favor of the creation of robust emotion classifiers for real-life applications.
Computer Speech and Language. Article in press.
Björn Schuller, Anton Batliner: "Computational Paralinguistics: Emotion, Affect and Personality in Speech and Language Processing", to appear, Wiley, ISBN: 978-1119971368, 2013.
if pitch is related with vocal folds and formants are with vocal tract, then how it occurs while taking FFT for magnitude plot of speech , Pitch and harmonics gets added. How can we insure that the formant in LP method is not a Pitch frequency?
Good question, Akhilesh ! Actually, in the standard speech production model, the vocal fold vibrations become the excitation (input) for the linear system, whose impulse response is defined by the particular configuration of the vocal tract at that time (for the limited duration of the current phone). Thus, these two signals convolve in the time domain and hence multiply each other in the frequency domain. So, the harmonics of the pitch very well show up in the speech spectrum.
In fact, due to the effect of the harmonics, usually, a formant gets aligned to the nearest pitch harmonic. However, in the linear prediction based spectral estimation, since only correlation at very low time delays are taken into account, the pitch harmonics disappear and you see only the formant peaks.
Can a same person can produce the same word in different pitch?
Yes, absolutely. Musicians can do it with ease. We can all do it with some effort. When a male person becomes red with anger and speaks, the pitch increases heavily, and you might even hear it as if a woman speaks !
sir,what is need of preemphasis in Lp analysis and is this equivalent to differentiating the speech signal 'diff ' function in matlab before appling Lp analysis
Actually sir when i am plotting LP analysis for vowel /a/, i am getting peaks at around 200,700.1200.2400 without diff function,but when i am using diff function i got 700,1200,2400. i can understand that diff function is removing pitch information. But if same processor is applied to hypernasal case which consistant of extra formant around 250 is also removed by diff function ie by preemphasis is we assume both are same.
what to do sir?
Dear Akhilesh,
You must read the basics thoroughly first. Preemphasis is used to boost higher frequencies in frequency modulation of audio in TV transmission. In LP analysis, the optimization criterion used only preserves peaks and not valleys. Since the higher formants are way below the first formants in energy, preemphasis ensures that the higher formants also are captured well by the LP spectrum.
Further, you can decide how many peaks (and hence, how many formants) are captured in the LP spectrum by changing the model order (p) of prediction.
Just got this link..may be useful
https://underlingsosu.wordpress.com/2013/03/08/phonetics-phriday-fundamental-frequency-harmonics-and-formant-frequencies/
In whispered speech, vocal folds do not vibrate and hence, there is no pitch. However, formants are still there, related to the phones being uttered.
Also, musicians vary their pitch wildly when they sing. When someone is "wildly" angry, the pitch increases tremendously. For example, the pitch of a man shouting in anger, can actually reach values close to those of women !
In case of voiced sounds, the air from the lungs vibrate the vocal folds. The fundamental frequency of vibrating vocal folds is the pitch. Now, the vocal fold vibration forces the vocal tract to vibrate as well. This forced vibration is resonance otherwise called as formant frequency. For voiced sounds, pitch can be estimated by autocorrelation method. The second highest peak is the pitch. Format frequency can be estimated by homomorphic filtering( In this we can separate the source of vibration(vocal folds) and vocal tract vibration).
I find this link very useful to understand the difference between pitch and formant frequency:
https://underlingsosu.wordpress.com/tag/formant-frequency/
Sound researchers in my institute have a spin-off company with aggression detection as an application, using harmonic-complex analysis.
http://www.soundintel.com/uploads/pdf/UK/Sound%20Intelligence%20Brochure%20(EN).pdf
http://www.ai.rug.nl/senscog/documents/van%20Hengel_2007.pdf
http://www.gavrila.net/Publications/cviu16.pdf
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing