What good is an emergency siren to an automatic driver assistance system (ADAS) that can’t hear? Radar, lidar, and cameras all rely on line-of-sight vision in some part of the spectrum.
But that provides no detection of objects that are obscured by obstacles, as commonly happens at blind intersections, where the sound of an approaching emergency vehicle or high-speed driver can provide a warning of out-of-sight hazards.
Detected sounds are displayed to to the driver using the head-up display to indicate an oncoming vehicle has been detected and its direction.
If ADAS systems are going to augment drivers’ senses, it needs to employ more of them. In this case, Infineon Technologies AG has developed a microphone system for cars that gives the ADAS a sense of hearing so it can detect unseen, and unseeable, threats.
This lets ADAS systems provide drivers with earlier notification of these threats and increase their safety margin.
Infineon's new sensing solution uses the company’s XENSIV MEMS microphones, AURIX microcontrollers, and Reality AI’s Automotive See-With-Sound (SWS) system. Reality AI’s machine learning-based algorithms let the system detect unseen emergency vehicles, cars, and other road participants.
A specific feature built into the system is the ability to distinguish the unique sounds of different countries’ emergency vehicle sirens. But there is no word on whether it provides any suitable commentary upon hearing, say, France’s distinctive flip-flop siren sound. Sacre bleu!
On the hardware side, the automotive-qualified XENSIV MEMS microphone has an increased operating temperature range from -40 °C to +105 °C allowing various use cases in harsh automotive environments. The low distortions (THD) and the high acoustic overload point (AOP) of 130 dB SPL enable the microphone to capture distortion-free audio signals in loud environments. These characteristics allow reliable classification of noises, even if the siren sound is hidden in high background or wind noise.
The sound from the mics needs to be processed, so the Reality AI software runs on Infineon’s AURIX TC3x processors, which are widely used across multiple automotive applications. The product range stretches from one to six cores and up to 16 MB of Flash.