What is the connection between voice interaction and smart home? Where are the technical barriers to voice interaction?

Imagine that when you come home from get off work and gently say "I'm back" to the door, the door not only opens automatically, but also turns on the air conditioner and TV within a short time after receiving your message. Is this feeling great!

In the future life described in many surgical phantom movies in many countries, the use of voice to control home appliances is also ubiquitous. In the "Iron Man" series of movies, after the actor returns home, he says he wants to drink coffee, and the coffee machine starts to make coffee. This kind of life is really enviable. With the continuous maturity of voice interaction technology, the pictures in future movies will soon become reality!

Voice interaction and smart home

Data show that in 2018, the scale of China's smart home will reach 180 billion yuan, and by 2020, the scale of the smart home market will reach 357.6 billion yuan. Analysts predict that the global smart home market will reach more than 500 billion yuan in 2021.

In the fields of automobiles and the Internet, voice interaction functions have become popular. Ford's SYNC system is a Ford in-vehicle multimedia communication entertainment system specially equipped for mobile phones and digital media players. It is a successful case of using voice interaction technology in current in-vehicle systems and has been widely used in many Ford series of cars. After Internet giant Apple launched its smart voice assistant application Siri in its iPhone 4S, Google also launched GoogleNow smart voice search and Q&A services in its Android smartphone operating system, and Microsoft also applied voice technology to Windows Phone.

In the field of smart home, foreign IT giants have successively entered the field of smart homes by combining smart home products and voice: Google acquired NEST to deploy smart homes and continuously strengthened the voice portal of Google Now; Apple's HomeKit smart home platform and Siri have also continued Strengthen integration; Microsoft also released the voice assistant Cortana (Xiao Na) to expand its interactive portal in the field of smart homes.

In China, as early as August 2014, the voice giant iFLYTEK announced its entry into the smart home market and released the intelligent voice assistant Lingxi 3.0 to control smart home devices. These devices need to be docked with Lingxi 3.0 beforehand. Search for the device in the App, and then perform voice control. There are already many devices supported, including TVs, coffee machines, electric lights, air conditioners, water heaters, etc.

With the increase of domestic and foreign giants' investment in the field of voice interaction, the core voice technology is gradually maturing, and the technical bottleneck that was once is slowly being broken. Among them, voice recognition technology is the foundation and core of voice interaction. "Voice recognition" technology is equivalent to installing "ears" in a computer system to enable it to have the function of "hearing". This technology undergoes complex steps such as speech signal processing, speech feature processing, model training and decoding engine, so that the machine can finally recognize the content, speaker, language and other information in the speech.

What is the connection between voice interaction and smart home? Where are the technical barriers to voice interaction?

Where are the technical barriers to voice interaction?

The previous article talked about the application of voice interaction technology in the field of smart homes. As one of the mainstream human-computer interaction methods, the voice interaction method frees people's hands from the touch screen and reduces the time people spend on data input. However, voice interaction is not suitable for all scenarios. At present, in the field of smart home, voice interaction still has the following problems:

What is the connection between voice interaction and smart home? Where are the technical barriers to voice interaction?

The recognition accuracy under far-field and noise interference needs to be improved and the noise reduction processing technology in endpoint detection. This is also known as the "cocktail party problem", meaning that it accurately recognizes a certain sound from a long distance in a noisy cocktail party. The current solution to this problem is multi-channel signal processing, such as the conversion of a microphone array. Relevant floor-to-ceiling technologies include the "Ring 6+1" microphone array of Spitz & Amazon Echo, and the optional installation voice pickup technology of the Youxiang Acoustic Mic.

Model optimization during endpoint detection, feature extraction and decoding. It may be simpler and more efficient if you put aside the assumptions and related designs of the existing problems and entrust them to the machine, and let the conversion model learned from the training data to convert the speech into a text sequence.

Currently, the end-to-end CTC model + Attention model is mainly used for optimization. Horizon’s internal evaluation proves that on 1000 hours of data, the performance of the CLDNN+CTC model is about 15% to 20% higher than that of the previous DCNN model. However, the end-to-end practicality is controversial. At this stage, it is only a step simplification of a certain part of the process, and the decoding part has not been included, and a large number of training sets are required.

Enhance predictability and adaptability. In daily interpersonal communication, people predict what the other person will say in the next sentence. The breakthrough for machines to obtain similar predictive capabilities lies in semi-supervised and unsupervised learning, especially reinforcement learning and transfer learning. Tencent previously had a PAC-RNN model that can adapt very quickly to continuously improve the recognition results. However, this model is more difficult to train due to the larger loop of the recurrent neural network.

In terms of hardware, chip development in the field of AI speech recognition is also a hot spot. In the terminal, the two key factors of voice recognition are real-time and cost. The development of dedicated chips for voice recognition is the development trend of terminal voice recognition hardware. Examples of terminal chips include: Qiying Tailun terminal intelligent voice recognition chip CI1006, Yunzhisheng UniRobot The hardware chip system, and the chip in the paper published by MIT on ISSCC2017.

summary:

In the field of consumption, smart home has changed the way consumers live. Voice interaction has made a lot of contributions to creating a safe, comfortable, convenient and information-based living space for smart homes, making people adapt to the fast pace of the information society. Maintain a completely open state of existence with the outside world. Smart home uses a family as a unit and uses a variety of information technologies to achieve the purpose of monitoring and information interaction. In the future, living will be smarter and consumers will have a more comfortable living experience. A voice interaction change seems imperative. .

Wired Gaming Mouse

Like to play games for the computer configuration,Gaming Keyboard have certain requirements,mouse as one of the most important accessories,is also very important.While Wireless Gaming Mouse is fine,there are still friends who find the delay and charging a hassle and prefer a wired mouse.Compared with the wireless mouse in business office portability. Wired mouse is more suitable for esports games,whether in terms of transmission delay or anti-interference ability convenient wired mouse is stronger than wireless mouse. Therefore, it is essential to choose a suitable wired mouse for those who often play esports games.


Playing games are generally selected cable mouse, cable mouse data transmission is stable,perfect,playing games do not have to worry about losing frames.The Wired Gaming Mouse has a huge advantage in stability because it is directly connected to the computer with a wire, so it has little interference from the outside world.More suitable for the mouse operation requirements of the game and design use. Wired mouse also has some disadvantages,such as can not operate too far, can only be used in the vicinity of the chassis (mouse wire length is limited).In addition,due to the cable, it will be a little messy feeling, and its disadvantages are summarized as follows. It drags and drags and feels uncomfortable to use, which is inevitable unless you have to use a wireless mouse.Use distance is short.This also can't be done.But if you don't think the mouse cable is troublesome, now there is a USB extension cord to sell, as long as you want.




T100 8

Wired Gaming Mouse,Led Wired Gaming Mouse,Wired Optical Gaming Mouse,Both Wired And Wireless Mouse

Henan Yijiao Trading Co., Ltd , https://www.yjusbcable.com