Recently, Xu Jing, director of qiniuyun Education industry, brought us a content sharing entitled “Live broadcasting 3.0 era, the fission and reconstruction of online education industry”. The following is a transcript of the speech.

Hello, everyone. My name is Xu Jing. As one of the earliest technicians in the live broadcasting industry, I have experienced all the eras of live broadcasting.

Evolution of live video

Live video has gone through three eras. The earliest live broadcast was actually a TV set, which played a standard definition signal, collectively called live broadcast. Many traditional media, such as Beijing TV and the former CCTV, are products of the era of live broadcasting 1.0. With the development of the Internet, live broadcasting entered the 2.0 era five or six years ago, which is marked by watching and hd. There were a lot of Internet startups, like Tencent Video, Youku, Tudou, iQiyi, that focused on high definition and viewability, but didn’t have any interaction. In the late stage of 2.0, simple interactions, such as tips, likes and bullet screens in the process of live streaming, gradually entered the market, which adapted to the simple interaction needs of the post-90s generation. Gradually, we found that all live streaming is valuable only when it is connected to interaction and communication. It is a social and communication between people, so that video gradually becomes a channel for sharing and expression.

At the same time, strong interaction means higher requirements for technology and lower latency.



Above is a data analysis by the ITU-T International Union. The horizontal axis is time, in milliseconds, and the vertical axis is user experience. It can be clearly found that real-time communication cannot be completed when the delay exceeds 400ms, and the delay exceeds 150ms, affecting user experience.

What are the key technologies to achieve latency below 150ms? Why wasn’t there such a low latency before 3.0? In the era of 1.0 and 2.0, RTMP is widely used to pursue high definition and precision, but TCP protocol is ignored as a security protocol. Three-way handshake, retransmission strategy and other security protocols ensure quality, while reducing the video transmission rate and improving the delay. Although UDP reduces the delay, it does not have a retransmission policy and relies heavily on the network. As shown in the triangle below, TCP and UDP have their pros and cons. How can Reliable UDP not only guarantee secure data but also take advantage of the ultra-low latency feature? The answer is Reliable UDP.



Reliable UDP combines the security of TCP and low latency of UDP to ensure video transmission through redundant coding and forward error correction, context-based retransmission policies, adaptive bandwidth adjustment, better congestion control algorithms, and multi-point relay. Based on these capabilities, WebRTC was launched, and the draft WebRTC 1.0 standard was published at the end of 2017 and officially released this year. More and more browsers and vendors began to support it extensively and build real-time audio and video applications based on WebRTC.



Trends in online education

Another important aspect of the educational scene is the whiteboard. A teacher of traditional education once lamented that many things in the traditional classroom were changing iteratively, but only one thing remained until now, that is the blackboard (whiteboard). This is enough to illustrate the importance of whiteboards in the context of education. Whiteboards solve online communication problems. For example, a colleague explained to me the difference between MCU and SFU in WebRTC, and he talked on the phone for more than ten minutes, but still had no idea what he was talking about. Then we plugged in the whiteboard demo (above), drew a few diagrams on the whiteboard, and in 2 minutes I got it all figured out. This is the magic of the online whiteboard, which makes online communication more efficient and modern education simpler and more intuitive.

Qiniuyun smart education solution also includes some very practical capabilities, such as Office components, which can use PPT or PDF function, so that modern education and traditional education can be combined to output education resources more quickly and effectively. Such as courseware sharing ability. Throughout the online education environment, teachers can upload pre-prepared documents so that students can directly download the teacher’s handout and lesson plan for review. Such as class recording, the entire learning process (teacher-student interaction, whiteboard interaction records, class notes) can be recorded for students to review after class.

In addition, Qiniuyun smart education solution integrates AI, auxiliary teaching to help teachers and students to teach more effectively. For example, in traditional education, the teacher speaks too fast, and the children can not take notes in class. Using speech recognition, the teacher’s speech is converted into words to record in real time, and text notes can be made at the end of the course. AI assisted teaching ability, let me give you another example. There is such a case in one-to-one teaching: a child said to an online foreign teacher, “Don’t take the class, I don’t want to take the class, this class is my father let me learn, you earn your money, I went to play games.” It turned out to be an ineffective class. This is where ai-assisted teaching comes in. Both ends of the camera will detect whether there is no face. If no face can be detected for more than one minute, the alarm mechanism will be triggered to inform the educational administration to follow up the learning process. The ai-assisted teaching ability also includes the number of hands raised by students who interacted with each other, with three students raising their hands 30 times in math class but sleeping in the row behind in English class. Then our AI-assisted teaching system will recommend Zhang SAN to take science courses. AI is to find students’ characteristics from a large number of low-density value data to assist students to make decisions.

Another is the application of big data in smart education. Traditional teaching records only semester grades, following students into the file. However, the data in the AI teaching system is collected, and each student has a large amount of data to follow, such as how many classes he or she skipped in college, which subject has special advantages in which period, etc., which will be recorded. In this way, the learning history of students can be more effectively reflected and the teaching files can be more complete.

Qiniuyun · Smart education industry solutions

Qniuyun smart Education solutions have corresponding solutions for the functions described above, from the most front-end collection coding SDK to CDN/VDN at the transport layer, to the core capability RTC network support, to the electronic whiteboard at the application end, recording at the back end, AI and big data analysis. The following enterprises can quickly access qiniuyun smart education solution to achieve low latency online education.

The first is the online education enterprise, now there are many online education institutions around us, especially in the past two years, the special fire, dada English, TopKid, even online teaching children piano, drawing and so on have also developed.

The second is double teacher education. For example, some American children communicate with some Chinese children under the guidance of one teacher, and these children are assisted by two local teachers. This is double teacher education. It’s hot in the US now and China will soon follow.

The third is vertical education. For example, long-distance medical teaching, mathematics teaching, programming teaching and so on. Another is corporate training, which can be very effective in improving the efficiency of corporate meetings.