How to discover the potential and high-quality anchors in the platform, recommend them to platform users, and create the next "live star" is a work that every live broadcast platform must consider. Effectively recommending high-quality anchors can not only improve the quality of the overall live broadcast room of the platform, but also attract more traffic from the outside to the platform through the anchor, and earn more traffic for the platform.
There are many live broadcast rooms in the current domestic live broadcast platform. How to find potential live broadcasters? What exactly can the "audio stream singing recognition function" bring to the platform?
In response to this scenario, Sumei Technology CEO Tang Huijun said: "As a leader in domestic AI online business risk control, we have been constantly enriching our products. This time, the custom-developed audio stream singing recognition function can detect the anchor in real time. Whether you are singing or not, feedback the results to the platform as a reference latitude recommended by the platform page. It will be very effective in improving the overall time retention of the platform."
What is the platform most afraid of? I was afraid that I would not be able to catch the trend of young people. What does the consumer market need? Flexible, efficient, personalized, "what I want, I want it now". In the face of the fiercely competitive live broadcast market, this challenge is even more obvious.
Generally speaking, in the scene of the webcast room, the sound is often diverse and complex: the sound may be the music played by the stereo in the room, the voice of the teammates in the game, or the anchor humming at will.
How to quickly select high-quality live broadcast rooms to maximize the effect of communication? Sumei Technology combines powerful voiceprint and audio scene detection technology, uses the VGG+RNN+Attention architecture to accurately locate the singing scene, and uses XVector+PLDA technology to intelligently verify the singing voice and the anchor's voiceprint automatically captured in the early stage , Remove interference from music and song playback, and provide real-time singing (live) recognition services. And by returning the recognition result every 10s, it can accurately identify whether the anchor is singing from the complex sound in the live broadcast room, and return the result to the live broadcast platform as a reference for priority ranking.
Through this function, the live broadcast platform can quickly screen out potential singing anchors from a large number of live broadcast rooms on the platform, and recommend and display them on the site.
Not only that, with the full-stack intelligent content recognition engine of Sumei Technology, it can also identify the pornographic, advertising, political and other sensitive information contained in the voice and video screens of the live broadcast room, effectively purify the live broadcast network environment, and actively respond to net-net operations. Provide a higher quality and better experience network environment for all users.