BioSonic is an intelligent, interactive web-based platform designed to visualize, process, and analyze diverse biological and acoustic signals.
By integrating signal processing, downsampling, and AI-powered diagnostics, BioSonic bridges the gap between medical signal interpretation and sound analysis—offering users a unified interface to explore ECG, EEG, Doppler, Drone, Voice, and SAR data.
With real-time visualization tools, multi-channel comparison, voice recording & recognition, anti-aliasing, downsampling support, and built-in machine learning models, BioSonic transforms complex waveforms into meaningful insights for both medical researchers and engineers.
BioSonic offers an end-to-end solution for visualizing and analyzing biomedical and acoustic signals.
It supports multi-channel viewing, AI-based classification, and dynamic 2D/3D representations for ECG, EEG, Drone, Voice, SAR, and Doppler signals.
The platform now includes:
- Automated downsampling for large biomedical datasets (EEG & ECG)
- Anti-aliasing filters for improved sound signal clarity (Drone, Doppler, and Voice)
- Voice recording with AI-based gender recognition
- Visualize multi-channel ECG signals in real time.
- Detect abnormalities such as arrhythmia, tachycardia, or fibrillation using an AI model.
- Supports automated downsampling for long ECG recordings to improve visualization performance without losing signal integrity.
- Multiple visualization modes:
- Continuous-time viewer (play, pause, zoom, pan)
- XOR Graph
- Polar Graph
- Reoccurrence Graph
Instructions:
- Upload your ECG dataset (
.csvor.mat). - Select the display mode and desired channels.
- Optionally apply downsampling to reduce data rate (choose factor 2×, 4×, etc.).
- The AI model automatically classifies signals as normal or abnormal.
- Use zoom/pan controls to explore specific signal segments.
- Analyze multi-channel EEG signals for neural activity patterns.
- Identify abnormalities such as seizures, epileptic events, or sleep disorders.
- Includes downsampling to handle high-frequency EEG recordings efficiently while preserving major waveform characteristics.
- Visualize in continuous, polar, or reoccurrence graph formats.
- Uses a 2D CNN model for automatic classification.
Instructions:
- Load EEG data.
- Apply downsampling if the signal has high sampling frequency.
- Choose the desired viewing mode.
- Adjust colormaps for better contrast.
- View real-time AI predictions for detected abnormalities.
- Detect and classify drone or submarine sounds within noisy environments.
- Employs an AI detection model for real-time sound recognition.
- Integrated anti-aliasing filter for removing high-frequency artifacts, enhancing detection accuracy.
Instructions:
- Upload or record an audio input.
- Apply anti-aliasing for clearer frequency response.
- View the spectrogram and activate drone detection mode.
- The model highlights identified drone sound patterns with detection confidence.
- Classify voice recordings to determine speaker gender (male/female).
- Accepts both uploaded audio files (
.mp3,.wav) and live voice recordings through the browser. - Integrated anti-aliasing filter for noise-free speech capture.
- Uses an AI-based model that analyzes MFCC, pitch, and spectral energy features to predict gender with high accuracy.
- Built-in waveform and mel-spectrogram visualizers for real-time feedback.
- Supports live microphone recording, enabling users to test in real time through their browser.
Instructions:
- Choose to record your voice directly or upload a pre-recorded file.
- Apply anti-aliasing to smooth out signal distortions.
- The system extracts key audio features automatically.
- The AI model classifies the input as Male or Female with confidence score.
- Visualize waveform and mel-spectrogram in real time.
- Visualize radiofrequency SAR signals and analyze spatial intensity variations.
- Detect motion or pattern-based characteristics from the data.
- Includes customizable 2D map representations.
Instructions:
- Upload SAR or cosmic signal data.
- Adjust visualization parameters.
- Observe AI-estimated signal features and distributions.
- Simulate a vehicle-passing Doppler effect using adjustable velocity (v) and frequency (f).
- Compare synthetic and real-world audio using AI to estimate vehicle speed and sound frequency.
- Integrated anti-aliasing to produce smoother audio transitions when generating synthetic Doppler shifts.
Instructions:
- Input vehicle speed and horn frequency.
- Enable anti-aliasing for higher quality Doppler synthesis.
- Generate and play the Doppler sound.
- Optionally, upload real audio to estimate its parameters.
If a model for any reason failed to download, just paste the files from manual download links in Server/ and unzip them.
Here are the manual download links:
| Folder | Google Drive Link |
|---|---|
| models | models.zip |
| EEG | EEG.zip |
| ECG | ECG.zip |
| ECG_XOR | ECG_XOR.zip |
| checkpoints | checkpoints.zip |
| Model / Dataset Name | Link | Short Description |
|---|---|---|
| PTB-XL Dataset | PhysioNet – PTB-XL | Large 12-lead ECG dataset for cardiac abnormality detection and downsampling experiments. |
| HMS Harmful Brain Activity Classification | Kaggle – HMS EEG | EEG dataset for detecting harmful brain activity, seizures, and evaluating downsampling efficiency. |
| Voice Gender Classification Dataset | Kaggle – Gender Recognition by Voice | Dataset of human voice recordings (male/female) used for AI-based gender classification and real-time recognition with anti-aliasing. |
| Vehicle Speed & Frequency Estimation | Hugging Face – Model | Regression model estimating vehicle speed and frequency from Doppler audio, enhanced by anti-aliasing. |
| Drone Audio Detection | Hugging Face – Drone Audio | AST-based binary classifier detecting drone sounds with integrated anti-aliasing for clean frequency filtering. |
| YAMNet | TensorFlow Hub – YAMNet | General-purpose sound classifier trained on AudioSet (521 classes). |
| SAR Detection (Synthetic Aperture Radar) | Alaska Satellite Facility (ASF) | Dual-polarization radar imagery for remote sensing and object detection. |
 |
 |
 |
 |
|---|---|---|---|
| Raghad Abdelhameed | Salma Ali | Youssef Mohamed Wanis | Rawan Mohamed |