Advanced Search:

Contact us

联系我们

Telephone:(852) 2838 3620

Email:sales@silverwing.com.hk

Address: Unit 2, 4/F, Kwai Cheong Centre, 50 Kwai Cheong Road, Kwai Chung, New Territories, Hong Kong

Silverwing and Ao Shuo Technology jointly explore human-computer interaction car voice

Source: Time:2018-11-06 11:00:22 views:

        2018 artificial intelligence first year, presumably AI is not an abstract concept. When AI landed in various vertical fields, such as home, car, transportation, education, etc., some products that were well known to the public came into being. For example, smart voice assistants, smart stereos, smart in-vehicle systems, and the like. The "AI Capabilities" platform has also been opened at home and abroad, allowing more developers and hardware vendors to share the "New Era" event. At the same time, artificial intelligence has also entered our daily lives “hot and vigorous”. Let's discuss "artificial intelligence" in the automotive field:

"Freeing hands" is no longer a legend!

        On October 19th, the 2018 China Automotive Networking International Summit hosted by Shanghai ECV (Europe) was successfully concluded in Shanghai. Mr. Zheng Tiansheng, the general manager of Aoshuo Tech, and Mr. Huang Shaozhong, the deputy general manager, attended the meeting and shared the future and challenges of human-computer interaction.

        Mr. Huang Shaozhong shared his speech on how to effectively help and improve the interactive experience in the car from years of experience in vehicle practical combat. Compared with the “smart technology” style of various technology companies in recent years, this theme is more like a deep-rooted, practical and pragmatic enterprise.

Voice development process and proud from "0" to "1"

        From Rex in the 1920s to the birth of the Hidden Markov Model (HMM) that affected the rapid development of speech in 1970, the HMM model officially introduced speech recognition technology in the 1980s. After nearly a century, in 2000 BMW introduced the first car with voice recognition. However, this car can only recognize 50 words, which can not meet the daily interaction needs of the owner. With the introduction of the Deep Neural Network (DNN) in 2010, deep learning technology has enabled speech recognition to have a qualitative fly in the car. Proudly launched in 2012, the first purely natural natural voice solution was introduced to the market in the first quarter of 2015. Proud Shuo Technology's "0" to "1" made up for the vacancy in the domestic market, and therefore won the order of the domestic No.1 depot.

Factors affecting the interactive experience in the car

1、the speech engine

        Today's speech recognition rate has reached 97%, 98%, but most car owners still say "machine can not understand words", "machine is so stupid" and other remarks. This is to mention the car driving scene, from the vehicle's own tire noise, car noise, the wind noise of the driving environment and the noise of the conversation inside the car. These factors directly affect the accuracy of speech recognition. Front-end acoustic processing is especially important in order for speech to recognize "clean" sound. The "SSE" module is a technology for solving interference sound sources, and improves the accuracy of speech recognition through techniques such as microphone array, sound source localization, beamforming, echo cancellation, and noise reduction.

        At the same time, there are difficulties in the core engines such as NLP and TTS. Mr. Huang Shaozhong also put forward profound insights, which will not be repeated here. In short, a stable and flexible voice architecture can handle the relationship between local and cloud relatively well, so that voice can be productized and users can get a good experience. It is a problem that we need to think about and solve.

2、Conflicts and contradictions between traditional car manufacturers and "internet thinking"

        It is well known that the car is an extremely complex and systematic product. However, with the innovation of automotive technology research and the higher demand of users for automobiles, the requirements of the vehicle manufacturers are constantly raised. In order to meet the needs of the development of users and communication intelligence, traditional car manufacturers have been compressing the cycle of cars in recent years. Empowering AI ability, car thinking is also in the contradiction, looking around and looking for ideas. However, the development logic of the Internet is dominated by product definitions. “I hope to be product-driven and user-driven”... But in the traditional car organization structure, how to make the product design as a whole under the existing framework Execution, this is conflicting! In the traditional architecture of the depot, the guarantee of time control and project quality is much larger than the so-called "bright spots" of the products.

3、HMI design

        There is no one in the car that interacts from the physical button to the GUI with the central control of the big screen and then to the current VUI. Decentralized UI design, so that each type of interaction is responsible for the corresponding part that can perform well, is the direction of product design. The voice does not need to start the engine, the voice needs to solve two or three steps of other interactions, and the car voice can be done in one step.

车载语言
Car voice interaction

                                          Home |  About us |  Product  |  Solution Provider  |   News |  Contact us  粤ICP备17091917号-1

                              HK Address: Unit 2, 4/F, Kwai Cheong Centre, 50 Kwai Cheong Road, Kwai Chung, New Territories, Hong Kong


Top