An Analysis of the Speech Recognition of Xiaomi TV 4A Core Technology

OFweek smart home network large-screen television traffic into the entrance, voice recognition into a core competitiveness. In recent years, the sales of large-screen smart TVs have continued to grow, and Internet TVs based on voice recognition have become more popular among consumers. Xiaomi TV 4A is one of them.

According to statistics published on the online retail sales of 2017W15 LCD TVs released by related data, millet TV sales continued to rank first in recent weeks, and stakeholders believe that the reason Xiaomi TV has secured its highest sales position in recent weeks has been recently released. The millet TV 4A is equipped with the artificial intelligence voice function.

Some people think that the biggest pain point of Xiaomi TV's 4A artificial intelligence voice recognition function is that the programs for the elderly and children are difficult to search, but they do not achieve 100% recognition in precise recognition technology. This is also the field of artificial intelligence voice recognition in the market today. A common problem.

Although the accuracy is less than 100%, in the smart TV industry, the voice recognition technology of Xiaomi TV 4A is still at the forefront.

Not long ago, Xiaomi Wangchuan had a live demonstration at the conference of Xiaomi TV 4A, which tested and interpreted the charm and deficiencies of artificial intelligence voice recognition technology from five levels respectively. Its main attraction features are the perfect implementation of the first four levels of testing: element-type search, fuzzy description search, chaotic character relationship sorting, and jump movies based on content, but use homonyms in the fifth level test. Errors in speech recognition were revealed when the instruction was incorrectly identified.

The picture shows Xiaochuan TV 4A in the recent press conference Wang Chuan successfully demonstrated the voice test link. Source: Network

Nowadays, most smart TVs have entered the field of artificial intelligence voice TV. Although there are differences in the precise realization of technical recognition, the technical logic is exactly the same.

Taking Xiaomi TV as an example, when we send a voice control command through the voice recognition system of Xiaomi TV 4A, the TV system collects the signal, and then converts it into a digital voice signal through its own analog data preprocessing, and then according to the module requirements. The digital speech signal data is sent to the cloud, processed via the cloud speech recognition analysis and sent to the cloud, so that we can control the TV to watch the desired media video content through voice.

Milli TV 4a's high-precision speech recognition and TV system module design is inseparable. It is understood that artificial intelligence voice TV system structure is divided into three modules.

Module 1: Speech Recognition System

This module mainly converts the analog voice signal collected by the TV into a digital voice signal in the cloud. The speech IC processing technology used in this stage can help the analog signal to be pre-processed, and then capture the feedback to the TV operating system through the feature parameters of the speech waveform.

The picture shows a small vocabulary word isolated speech recognition waveform in Chinese, from: Network

Module 2: TV System Processing

The television system is the bridge between the speech recognition system and the cloud processing system. In the state of access to the Internet, the television system receives speech and preprocesses it, and has a specific module feature for a specific television system, and transmits the module characteristics and voice data to the cloud.

Module 3: Cloud Processing System

As Ma Yun said: "People are not as good as days, and days are clouds." The cloud processing system analyzes and processes digital voice data through cloud computing, and performs cloud intelligent identification to complete the corresponding voice command function.

Analysts believe that the smart TV's speech recognition technology will still be the core competitiveness of many smart TV brands in the future. In addition, scene applications such as VR will also become important entrances for smart TVs.

Diesel Generating Set

Diesel Generating Set,Genset Generator,Independent Power Supply,Office Buildings Generator

Shaoxing AnFu Energy Equipment Co.Ltd , https://www.sxanfu.com