tags: ece-402 dsp spectral-analysis
Time-Varying Spectral Analysis
Time-Varying Spectral Analysis is a method of [spectral analysis] that generates a series of [amplitude] and [frequency] [envelope]s that describe a [sound].
- kind of an oxymoron: [Fourier transforms] already utilize time; you can't take a Fourier transform of just one sample
- trying to analyze sound as we perceive it: over a period of time
Time-varying analysis methods offer various degrees of information on the [partial]s within a [spectrum], including:
- partial amplitudes
- partial frequencies
- partial [phase]s
- in steady tones, phase does not matter (Ohm, 1843). Ex: [Hilbert Transformed Square Wave]
- However, in the [transient]s of noises, phase can drastically alter the [timbre] of a sound
- frequency is derivative of phase
- ex. glottal pulses in speech
Types
There are 2 main types of time-varying spectral analysis: