Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 11

Revision of DSP

Basic Audio Signal


Digitalization of Signal
Types of Signals
• Periodic
• Periodic signal with period N
• X[n+N] = x[n]
• Aperiodic
• When there is no repetition in the signal then its aperiodic
• There is no N where X[n+N] = x[n]
• Symmetric Signal
• X[-n] = X[n]
• Asymmetric Signal
• X[-n] = - X[n]
Types of Signals
• Stationary
• Signal doesn’t change its properties along the time
• Amplitude, frequency of the signal are constant wrt time
• X[n] = sin(theta)

• Non Stationary
• Signal changes properties with time

• NOTE: Audio is non stationary we need to make it stationary


Discrete Time System
• y[n] = 0.5 x[n] + 0.6 x[n-1] + 0.8 y[n-1]

• So output signal depends on


• Current input x[n]
• Previous input (input delayed by 1) x[n-1]
• Previous output (output delayed by 1) y[n-1]

• Such systems are possible in discrete time systems


Types of DTS
• Static / Memory less DTS
• When the output depends on
• current input
• not on history
• Dynamic / Memory System
• When output depends on
• Historical inputs
• (OR / AND) Future inputs
• Time In-variant System
• When output doesn’t vary with time
• Y[n] = H[x[n]] (always)
• Ex. Any computer program / hardware
• Time Variant System
• When output may change with time (ex. Human vocal cord )
• NOTE: Signal are generally non stationary but DTS systems are time in-variant
Types of DTS
• Causal System
• If system output only depends on current historical inputs
• Doesn’t depend on future
• Non causal
• If system depends on future inputs then its non causal
• Linear System
• H[x[n] + x[n-1]] = H[x[n]] + H[x[n-1]]
• It follows this suppression principle
• NOTE: A Linear Time Invariant system is LTI systems
• In Speech processing we consider a LTI system

• Recursive System:
• If current output depends on previous output
Convolution of 1D Signal
• input = [1, 0, 2, 3, 0, 1, 1] • input = [1, 0, 2, 3, 0, 1, 1]
• kernel = [2, 1, 3] • kernel = [2, 1, 3]
• Padding = no , stride =1 • Padding = Yes , stride =1
• Output = [8, 11, 7, 9, 4] • Output = [1,8, 11, 7, 9, 4,3]

1 0 2 3 0 1 1 0 1 0 2 3 0 1 1 0
2 1 3 2 1 3

1 0 2 3 0 1 1 0 1 0 2 3 0 1 1 0
2 1 3 2 1 3
1 0 2 3 0 1 1 0 1 0 2 3 0 1 1 0
2 1 3 2 1 3
Auto correlation
• Correlation between signal and its own delayed copy

• (A.k.a.) Serial correlation

• For Time series or audio signal

You might also like