Analysis Methods for Speech Sounds Flashcards
recording chain
microphone > cable > hardware > computer
microphones change air pressure changes into
analog electrical signals
quantization
the discretization of the signals amplitude
regular sampling rate
chosen when digitizing signals
signal aliasing
when signal sampled is a higher frequency then the nyquist frequency creates a “fake signal”
pure tones
sine waves
simplest wave
no timbre characteristic
complex signals
speech
still periodic
sum of many sine waves
have timbre characteristic
quasi-periodic
similar to complex periodic signals, except they are undergoing changes
Fourier transformation
represent the signal in terms of the frequencies that make up that signal
discrete fourier transformation
uses a fixed length of signal as an input, which makes the computation simpler
fast fourier transformation
algorithm that takes an input length that is always 2^n points long
fundamental frequency
is the greatest common denominator of all harmonics, and the duration of the period is the lowest common multiple
spectra
indicate which frequencies are present and how prominent they are in a signal
If the window does not match the period of the signal the reconstructed signal will be
distorted and may contain additional frequency components based on the period of the window
what part of the window do we care about most
middle
