Skip to content
View Moustafa-Afara's full-sized avatar
0 followers · 1 following

Block or report Moustafa-Afara

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Moustafa-Afara/README.md

Moustafa Afara — signal processing and pattern recognition (audio · biosignals · vision)

I take a signal — speech, muscle or brain activity, a video frame, a system state — extract the feature that carries the information, and classify or track it with a learned model. BSc in Informatics Engineering, artificial-intelligence specialisation (2020).

Research interests: information hiding and cryptography built on deep learning, computer vision and signal processing; cross-domain features (speech features on biosignals); honest evaluation — on people and sessions the model has never seen.

Projects

Domain Project Result (dataset)
Biosignals MFCC vs time-domain features for EMG gesture recognition 80.7 % on unseen people, 6 gestures (UCI, 36 people); MFCC +2.5 pts with LDA (p = 0.009), −2.5 with an MLP
Biosignals EMG-controlled prosthetic hand — controller + Arduino 84.4 % grip decisions offline on unseen people, ≈ 400 ms latency; not yet tested on hardware
Biosignals EEG sleep staging with CNN + LSTM — study critical analysis of a published pipeline's evaluation (subject leakage)
Audio Speaker identification with MFCC + GMM 98.0 % same-session, 81.1 % cross-session (LibriSpeech, 40 / 31 speakers)
Audio Music genre classification with MFCC + MLP 61.0 % over 10 genres, duplicates removed (GTZAN)
Vision Real-time multi-object tracking with YOLOv3 + OpenCV detection + IoU tracking with persistent IDs
Vision Traffic speed-violation demo on YOLOv5 tracking and labelling pipeline; speed input simulated
Control Linear and nonlinear MPC in MATLAB — study report and results

Every repository states where its data and any borrowed code come from, how it was evaluated, and what it does not show.

Next

  • Vehicle speed estimation from camera and microphone together (VS13 dataset).
  • An Arabic article on whether a speech feature (MFCC) works on muscle signals.

Writing

«الوصف الدقيق لخوارزميات التعلم الآلي والعميق وتطبيقها في الرؤية الحاسوبية» — a 99-page Arabic draft book (2021–2022) tracing object detection from Felzenszwalb segmentation through R-CNN, SSD and YOLO v1–v5, with the mathematics. Most projects above ship with an Arabic report.

Tools: Python (PyTorch, scikit-learn, OpenCV, librosa), MATLAB, Kotlin, Arduino · Arabic and English.

Popular repositories Loading

  1. yolov3-realtime-object-tracking-opencv yolov3-realtime-object-tracking-opencv Public

    Real-time multi-object detection and tracking with YOLOv3 on OpenCV DNN and an IoU tracker: persistent IDs, motion trails, webcam or video.

    Python

  2. speaker-identification-mfcc-gmm speaker-identification-mfcc-gmm Public

    Closed-set speaker identification with MFCC features and per-speaker GMMs, evaluated on LibriSpeech (98.0% any-session, 81.1% cross-session).

    Python

  3. mpc-linear-nonlinear-matlab-study mpc-linear-nonlinear-matlab-study Public

    Study of linear and nonlinear model predictive control in MATLAB (Arabic report + result), crediting M. T. Augustine's MPC-MATLAB programs.

  4. eeg-sleep-staging-cnn-lstm-study eeg-sleep-staging-cnn-lstm-study Public

    Study and critical analysis of a CNN + LSTM sleep-stage classifier on single-channel EEG (Sleep-EDF): method, reported results, and evaluation pitfalls.

  5. emg-gesture-recognition-mfcc-vs-td-features emg-gesture-recognition-mfcc-vs-td-features Public

    Does a speech feature work on muscle signals? MFCC vs time-domain features for 6-gesture sEMG recognition, 36 people, tested on unseen subjects with paired statistics.

    Python

  6. emg-prosthetic-hand-prototype-ann-arduino emg-prosthetic-hand-prototype-ann-arduino Public

    EMG-controlled 5-finger prosthetic hand: TD + MLP grip controller (84% on unseen people, offline), Arduino servo receiver, cleaned 2020 MATLAB prototype. Not tested on hardware.

    Python