Sensing and Deep Learning

The ADAS and AD market size is expected to reach ~USD 43 billion by 2030 for major software development efforts. Driver Monitoring Systems/Occupant Monitoring Systems/Front Fcacing Camera are considered essential due to regulatory requirements such as EU NCAP 2022.  In addition, OEMs are also moving towards L4/L5 ADAS features at the same time. Therefore, the challenges for most OEMs are getting the right set of training data for algorithm development, finding the right talent pool, and getting higher accuracy and performance with lower hardware resources. OEMs also face the challenges in complying with safety requirements across multiple geographies including the establishment of integration labs for testing. Selecting the right partner can solve many of these challenges with expertise, IP, tools and labs. 

HARMAN has strong capabilities in ADAS, Machine Learning and Deep Learning to help build solutions that solve current and future OEMs ADAS roadmaps. HARMAN can help in accelerated ADAS algorithm development, integration on different SoC and long-term maintenance & support. HARMAN also has ADAS testing labs available to OEMs providing HiL, MiL and SiL testing with the help of our own proprietary toolchain. We have a large team located in the US, Poland, India and Russia specialized in ADAS algorithm development, testing and integration. Expertise covers deep learning, computer vision, embedded solution, algorithm optimization and performance improvements for sensors like Camera and LiDAR. We also have proficiency in:

• Texas Instruments – Texas Instruments Deep Learning (TIDL)
• Qualcomm – Neural Processing SDK for AI (SNPE)
• NVIDIA – NVIDIA Deep Learning Accelerator (NVDLA) and AGX Platform

HARMAN ADAS Practice has products which can act as solution accelerators. Our four product streams are:

• Interior Sensing (CMS/DMS) – Cabin and Driver monitoring systems. These are productionized and hardened algorithms developed in modular fashion and provide 5 features (including Object Detection) and 12+ solution accelerators (including Head Pose + Gaze Direction). 
• Exterior Sensing (FFC) – Front Facing camera use cases are also becoming a commodity. HARMAN provides core algorithms as an enabler for advance functionality. The market is moving into areas such as rearward alerts, traffic signs and emergency vehicles. Our algorithm has 15+ features (including Traffic light Phase Detection) that are supported or planned in the roadmap. 
• ADAS for Motorcycles – ADAS for High End Motorcycles are introducing new safety features. HARMAN can see strong demand from all geographies. We have 5 features (FCW, HBA, Traffic sign recognition, etc) available and an additional seven features in roadmap.
• Sound Recognition – HARMAN is using AI for small chips and scratches on the vehicle. This would reduce the cost compared to using other mechanisms. This is essential for rental car fleets and could be integrated into OEMs core platform for alerts, much like theft. In addition HARMAN leverages their strong Car Audio capabilities for microphone design.