Chongqing- Changan Automobile recently unveiled the world's first out-of-car voice interaction system. Based on the deep integration of AI technologies and car usage scenarios, this intelligent system is the brainchild of a young team from the Chongqing Branch of iFLYTEK Intelligent Automobile.

iFLYTEK set up its Southwest Headquarters in Chongqing in 2018. It has also brought the Chongqing Branch of iFLYTEK Intelligent Automobile. iFLYTEK cooperated with Changan Automobile and achieved multiple technological breakthroughs in the past years.

A huge scenario-related knowledge base

"Hello, Xiao'an. Please check the weather in Beijing for me."

"Ok. It is sunny in Beijing with a temperature of 20°C in the daytime."

"Book a flight there for me."

Such dialogues are very common for people, but it is not easy for human-machine interaction to keep a conversation going and let a machine understand that "there" refers to Beijing.

"People sitting in the car only need to awaken the in-car system once to let the system carry multiple instructions. The system can understand instructions in context. Such kind of interaction has never been realized before," said Ma Hongpeng, the general manager of the Chongqing Branch of iFLYTEK Intelligent Automobile.

Ma Hongpeng, the general manager of the Chongqing Branch of iFLYTEK Intelligent Automobile, instructed his colleagues to test the intelligent vehicle system. (Photo provided to iChongqing)

"The machine needs to memorize and understand the speaker's intentions and the scenarios, so it is necessary to include as many scenarios as possible. The algorithm will become exponentially difficult, resulting in difficult software development and maintenance," Ma added.

After many tests, Ma and his team successfully built a vast scenario knowledge base. Now, iFLYTEK's Flying Fish operating system (OS) has been applied to CS95, CS75, and other Changan cars.

The speech recognition system goes local

As many Chongqing auto brands have started to go global, creating the in-car system in foreign languages is necessary. 

An excellent in-car speech recognition system should understand a foreign language, cultural habits, and accents. So the system needs a powerful recognition model and learning ability. 

Ma and his team divided the in-car speech recognition into two technical levels. One is "voice recognition," which is like a human's mouth and ears and involves expressing and obtaining information; the other is named "semantic recognition," which is like a human's brain and takes control of thinking and processing information. 

In 2021, the in-car foreign-language speech recognition system was invented, and all Changan Automobile's models launched overseas have adopted this system.

Awaken the car from outside

"In the past, the voice interaction system was applied inside the car. The intelligent interaction outside the car is also a user's demand," said Ma. 

The out-of-car voice interaction system enables the owner to awaken the car outside. For example, the vehicle can move forward in a crowded parking lot when saying "drive out." When you return from shopping with bags in your hands, the trunk will open if you say "open the trunk."

The team has created two technologies to carries out instructions with external noise. One is the neural network noise reduction algorithm, and the other is named the "cold start" system. In this way, the system can achieve 360-degree noise reduction and recognition coverage.

Intelligence, networking, and sharing have defined the second half of the car revolution. More robust and innovative intelligent scenario services will show up in Chongqing's auto ecosystem and facilitate Chongqing's smart car manufacturing to a new level.