An academic investigation into the temporal and spectral properties of a discrete-time Wide Sense Stationary (WSS) Gaussian random process modeled as a first-order autoregressive, or AR(1), system.
- Author: Aakash Nair (Reg. No: 24BEC0080)
- Course: Random Processes
- Faculty Advisor: Dr. Mugelan R K
- Tools: MATLAB & Python (NumPy, SciPy, Matplotlib)
This project provides complete mathematical formulation, empirical simulation, and statistical verification of an AR(1) Gaussian random process. The empirical estimates of mean, autocorrelation, and Power Spectral Density (PSD) are compared with theoretical closed-form expressions to validate wide-sense stationarity and spectral shaping properties.
The first-order autoregressive process
where:
-
$a = 0.8$ is the autoregression coefficient. The stability and WSS condition requires$|a| < 1$ . -
$W[n] \sim \mathcal{N}(0, \sigma_w^2)$ is discrete-time White Gaussian Noise with variance$\sigma_w^2 = 1.0$ .
-
Mean:
$$\mathbb{E}[X[n]] = 0$$ -
Autocorrelation Sequence:
$$R_X[k] = \mathbb{E}[X[n]X[n-k]] = \frac{\sigma_w^2}{1 - a^2} a^{|k|}$$ -
Process Variance (Zero-lag Autocorrelation):
$$\sigma_X^2 = R_X[0] = \frac{\sigma_w^2}{1 - a^2} = \frac{1.0}{1 - (0.8)^2} = \frac{1.0}{0.36} \approx 2.7778$$ -
Power Spectral Density (DTFT of
$R_X[k]$ ):$$S_X(e^{j\omega}) = \frac{\sigma_w^2}{|1 - a e^{-j\omega}|^2} = \frac{\sigma_w^2}{1 - 2a \cos(\omega) + a^2}$$ - Low-frequency peak (
$\omega = 0$ ):$S_X(e^{j0}) = \frac{1.0}{(1-0.8)^2} = 25.0$ - High-frequency roll-off (
$\omega = \pm\pi$ ):$S_X(e^{j\pi}) = \frac{1.0}{(1+0.8)^2} \approx 0.3086$
- Low-frequency peak (
Shows the first 200 samples of the synthesized random sequence oscillating about zero mean with short-term correlation.
Empirical autocorrelation sequence computed via biased cross-correlation against the theoretical exponential curve for
Empirical PSD estimated using Welch's averaged modified periodogram compared against the analytical DTFT spectrum across
Complete multi-panel visualization generated via Python simulation pipeline:
| Metric | Theoretical | Empirical (Simulated) | Absolute Error |
|---|---|---|---|
| Mean |
|||
| Variance |
|||
| Zero-Lag |
|||
| Autocorrelation MSE | |||
| Peak PSD |
|||
| Roll-off PSD |
Open MATLAB and execute:
run('project.m')This runs the simulation, generates Figures 1, 2, and 3, and exports the high-resolution figures.
- Install dependencies:
pip install -r requirements.txt
- Execute the simulation script:
python simulate_ar1.py
This prints the statistical validation table and updates the figures in figures/.
├── README.md # Project overview, theoretical derivation, and results
├── LICENSE # MIT License
├── requirements.txt # Python dependencies
├── .gitignore # MATLAB, Python, and OS ignore rules
├── project.m # MATLAB simulation and figure export script
├── simulate_ar1.py # Python simulation script with statistical reporting
├── docs/ # Documentation and report files
│ ├── report.md # Detailed academic report
│ └── RP_Project.docx # Original docx project report
└── figures/ # High-resolution simulation plots
├── fig1_time_domain.png
├── fig2_autocorrelation.png
├── fig3_power_spectral_density.png
└── fig_combined_analysis.png
This project is licensed under the MIT License - see the LICENSE file for details.



