Siren detection for autonomous cars using Machine Learning

Client Overview

US-based technology company providing next-generation solutions for automotive industry.

Business Challenge

The client wanted to develop a feature for autonomous cars that will intelligently identity ambulance/fire trucks/police siren and will give way to it. For the same, client required a development partner with expertise in ML in automotive sector who can help them develop the functionality accelerating their launch timelines.

VOLANSYS Contribution

Signal Processing Algorithms
- Designed Band Pass Filter to remove unwanted frequencies
- MFCCs computation to extract the audio features
- Fourier Decomposition to apply statistics on sub bands
Machine Learning – Deep Learning
- Designed the solution using supervised learning algorithm–Artificial Neural Network (ANN)
- Developed model using TensorFlow framework and trained model with extracted features from 9000 audio files
- Trained model has 90% prediction accuracy
- Developed application on NXP i.MXRT series for extracting the features from real audio samples, pre-processing and running inference for ambulance siren detection
Quality Engineering
- Performance evaluation using real ambulance siren through microphone

Technologies | Engineering Expertise

Machine Learning | Deep Learning | Quality Engineering | TensorFlow | Signal Processing

Application Diagram

Ambulance siren detection diagram

Benefits Delivered

Improved on-road safety with audio detection feature implementation
Helped upsell end-customer’s product by enhancing the autonomous car feature set

Ambulance Siren Detection for Autonomous Cars using Machine Learning