FUNDAMENTALS OF SIGNAL PROCESSING
1. Signals: Mathematical Representation
• Formal Definition: A signal x(t) is a function of an independent variable t (typically
time for 1D signals). For multidimensional signals (like images), we might have
x(t, s) or x(x, y).
• Continuous-Time Signals: Represented by functions defined over a continuous
range of t. Examples:
o Sinusoidal Signal: x(t) = Acos(2πf₀t + φ), where A is amplitude, f₀ is
frequency, and φ is phase.
o Exponential Signal: x(t) = Ae^(αt), where A is amplitude and α is a
complex exponent.
• Discrete-Time Signals: Represented by sequences x[n], where n is an integer
representing the sample index. They are obtained by sampling continuous-time
signals: x[n] = x(nT), where T is the sampling period.
• Key Mathematical Operations on Signals:
o Addition: z(t) = x(t) + y(t)
o Multiplication: z(t) = x(t)y(t)
o Scaling: z(t) = ax(t), where a is a scalar.
o Time Shifting: z(t) = x(t - τ), where τ is the time shift.
o Time Scaling: z(t) = x(at), where a is a scaling factor.
2. Spectra: The Fourier Transform and Frequency Domain
• Fourier Transform (FT): A powerful mathematical tool that decomposes a signal
into its constituent frequencies.
o Continuous-Time Fourier Transform (CTFT):
§ X(f) = ∫₋∞⁺∞ x(t)e^(-j2πft) dt
§ x(t) = ∫₋∞⁺∞ X(f)e^(j2πft) df (Inverse CTFT) Where j is the imaginary
unit, and f represents frequency.
o Discrete-Time Fourier Transform (DTFT):
§ X(ω) = Σₙ₋∞⁺∞ x[n]e^(-jωn)
§ x[n] = (1/2π)∫₋π⁺π X(ω)e^(jωn) dω (Inverse DTFT) Where ω is the
angular frequency.
o Discrete Fourier Transform (DFT): Used for digital signals of finite length.
It transforms a sequence of N complex numbers x[0], x[1], ..., x[N-1] into a
sequence of N complex numbers X[0], X[1], ..., X[N-1]:
§ X[k] = Σₙ₀ᴺ⁻¹ x[n]e^(-j2πkn/N)
• Properties of the Fourier Transform:
o Linearity: The FT of a linear combination of signals is the linear
combination of their FTs.
o Time Shifting: A time shift in the time domain corresponds to a phase
shift in the frequency domain.
o Time Scaling: Time scaling in the time domain corresponds to inverse
scaling in the frequency domain.
o Convolution Theorem: Convolution in the time domain corresponds to
multiplication in the frequency domain, and vice-versa.
• Power Spectral Density (PSD): Describes how the power of a signal is
distributed over different frequencies. For a signal x(t) with Fourier Transform
X(f), the PSD is often defined as Sₓₓ(f) = |X(f)|².
3. Signal Processing: Mathematical Operations and Algorithms
• Convolution:
o Continuous-Time: (x * y)(t) = ∫₋∞⁺∞ x(τ)y(t - τ) dτ
o Discrete-Time: (x * y)[n] = Σₖ₋∞⁺∞ x[k]y[n - k]
• Filtering: Implemented using convolution or multiplication in the frequency
domain. Filter design involves determining the appropriate filter coefficients or
frequency response.
• Z-Transform: A generalization of the DTFT that is useful for analyzing and
designing discrete-time systems.
• Sampling Theorem (Nyquist-Shannon): States that a bandlimited signal (a signal
with no frequency components above a certain frequency B) can be perfectly
reconstructed from its samples if the sampling rate fₛ is greater than twice the
highest frequency component (fₛ > 2B).
• Digital Filtering: Implemented using difference equations or convolution with a
finite impulse response (FIR) filter or an infinite impulse response (IIR) filter.
• Correlation: Measures the similarity between two signals.
o Cross-correlation: Rₓᵧ(τ) = ∫₋∞⁺∞ x(t)y(t - τ) dt
o Autocorrelation: Rₓₓ(τ) = ∫₋∞⁺∞ x(t)x(t - τ) dt
