TU-TESL

• 22

According to foreign media reports, now, pedestrians with headphones or earphones can be seen everywhere in the streets of big cities. They have been staring at mobile phones while walking, completely ignoring the surrounding environment, i.e. "low headed people" (playing with mobile phones while walking), and usually bring danger. Pedestrians wearing earphones often do not hear audible warning of injury, such as the sound of horns, shouting or approaching vehicles. As a result, in the past seven years, the number of casualties caused by "playing mobile while walking" in the United States has tripled, and the number of pedestrian deaths in the United States last year reached the highest level since 1990. In response to this growing public safety concern, researchers at the data Science Institute (Columbia) at Columbia University are designing a smart headset system that can alert pedestrians of danger. Equipped with micro microphone and intelligent signal processing device, the headset can detect the sound of approaching vehicles. If there is a danger nearby, the system will give an audible alarm to the headphones of pedestrians. The team is developing prototype headphones and testing them on the streets near Columbia University. Once successfully developed, the intelligent wearable system can help reduce pedestrian casualties. In 2017, the smart headset project was supported by the National Science Foundation of China (NSFC) for us $1.2 million (about 8.4 million yuan). Since then, the team has also published two conference papers and one journal paper in the IEEE Internet of things journal. The development of this smart headset is very complex. It requires the integration of various micro microphones into the headset. It also requires the development of low-power data channels to deal with all sounds near pedestrians. It must also be able to extract the correct signals to indicate the imminent danger. The data pipeline will include an ultra-low power consumption custom integrated circuit, which can extract relevant features from the sound without consuming power. In addition, the researchers are also using the most advanced data science and technology to design the smart headset. The machine learning module on the user's smartphone can classify hundreds of sound signals emitted by the city streets and nearby vehicles, and give an alarm when the user is in danger. Researchers will design mechanisms to enable people to identify alarms and respond quickly. At present, the team is testing the design in the laboratory and on the streets of New York City, which is famous for its congestion and noise. Psychology professor at Barnard College at Columbia University says people will be experimented with perception and behavior to see how to effectively alert pedestrians walking in the city with headphones. The goal is to develop a prototype smart headset system at Columbia University and then transfer the technology to a commercial company, the researchers said.