The North Amercian division of a Canada-based automotive parts supplier was looking for a low power consuming solution for OD (Object Detection) and PD (Pedestrian Detection) algorithms for their ADAS (Advanced driver-assistance systems). Their existing algorithms were optimized for PC and hence were not suitable for deploying into automotive environment.

The solution required optimization of the algorithms and porting the same on a very low power consuming chipset of a car DSP (Digital Signal Processor). eInfochips was selected as a solution provider, considering the expertise on Broadcom platforms. The platform to be used was Hydra DSP, which though was a low power consuming platform, posed the constraints of extremely low data memory and instruction memory.

eInfochips successfully delivered the phase one of the project, which involved porting of camera modules in a very short time. This resulted in winning the complete project from the client. In the second phase of the project, eInfochips carried out conversions from C++ to C and floating point to fixed point, before porting of 4 ADAS cases along with audio mixing and processing for 3 audio algorithms.

The client benefited from the deployment of reusable algorithms optimized for porting on any chips in an automotive environment, consuming minimal power and hence making it suitable for upcoming generations of the traditional as well as electric vehicles.