MATLAB Projects in Bangalore for ECE

A complete OFDM transceiver is implemented in MATLAB simulating 64-subcarrier OFDM with QPSK, 16-QAM, and 64-QAM modulation, cyclic prefix insertion, and pilot-based channel estimation using the LS algorithm. The system is simulated over a 6-tap Rayleigh fading channel at Eb/N0 values from 0 to 30 dB. BER curves are plotted and compared against theoretical AWGN bounds. The simulation demonstrates the effectiveness of cyclic prefix in eliminating inter-symbol interference and shows that pilot spacing of 4 subcarriers achieves near-perfect channel estimation at SNR above 15 dB.

An adaptive noise cancellation system is implemented in MATLAB using both the Least Mean Squares (LMS) and Recursive Least Squares (RLS) algorithms. A clean speech signal is corrupted with additive white Gaussian noise and 50 Hz power-line interference. The adaptive filter converges to cancel the noise reference signal. Performance is evaluated using SNR improvement, convergence speed, and misadjustment. RLS achieves 18 dB SNR improvement in 50 iterations versus 200 iterations for LMS. Audio playback of the original, noisy, and filtered signals is included in the demonstration script.

A MATLAB implementation of Fuzzy C-Means (FCM) clustering segments brain tumour regions from T1-weighted MRI images sourced from the BraTS 2020 dataset. Preprocessing includes skull stripping, histogram equalisation, and Gaussian filtering. FCM with 4 clusters separates tumour, oedema, grey matter, and white matter regions. Segmentation accuracy is evaluated using Dice coefficient (0.84) and Jaccard index (0.73) against manual expert annotations. The project includes a MATLAB GUI that loads a DICOM image, runs segmentation, and overlays the tumour mask on the original scan.

A DC motor speed control system is modelled in Simulink using a second-order transfer function derived from motor datasheet parameters. A PID controller is designed using the Ziegler-Nichols tuning method and further optimised using the MATLAB PID Tuner app to achieve a settling time of 0.8 seconds and zero steady-state error. Step response, ramp response, and disturbance rejection are simulated and compared against the uncontrolled system. The Simulink model includes anti-windup integration limits and a rate limiter on the control output to simulate actuator saturation.

A MATLAB implementation embeds a 32×32 binary watermark into a host image by modifying mid-frequency DCT coefficients of 8×8 blocks. The watermark is spread using a pseudo-random sequence keyed to a secret key. Robustness is evaluated against JPEG compression (quality 50–90), Gaussian noise addition, median filtering, and geometric attacks (rotation, cropping). The normalised correlation between the extracted and original watermark is computed for each attack. At JPEG quality 70, the system achieves NC = 0.91, demonstrating robust watermark survival under common image processing operations.

A speaker-independent digit recognition system (0–9) is implemented in MATLAB using Mel-Frequency Cepstral Coefficients (MFCC) for feature extraction and Hidden Markov Models (HMM) for classification. The TIMIT digit subset with 100 speakers is used for training and testing. Each digit is modelled by a 5-state left-to-right HMM with 8 Gaussian mixture components trained using the Baum-Welch algorithm. The system achieves 94.3% word accuracy on the test set. A real-time demonstration records a spoken digit via the PC microphone and displays the recognised digit within 500 ms.

A MATLAB script analyses power quality in a simulated three-phase distribution system by computing the FFT of voltage and current waveforms sampled at 10 kHz. Total Harmonic Distortion (THD) is calculated for fundamental frequencies of 50 Hz with harmonics up to the 25th order. The simulation introduces non-linear loads (rectifier, variable-speed drive) and measures their harmonic contribution. Results are presented as harmonic spectrum bar charts and a THD trend plot over a 24-hour load cycle. The project demonstrates power quality analysis techniques per IEC 61000-4-7 standards.

A MATLAB implementation compresses grayscale images using a 2D Discrete Wavelet Transform (DWT) with Daubechies-4 wavelets followed by threshold-based coefficient quantisation and Huffman entropy coding. Compression ratios of 8:1 to 20:1 are achieved with PSNR values of 38–42 dB, outperforming JPEG at equivalent bit rates for smooth images. The project includes a comparison of Haar, db4, and sym8 wavelets at multiple decomposition levels. A MATLAB GUI allows users to load an image, select wavelet type and threshold, and view the original, compressed, and difference images side by side.