"AIoT" is the key word of our times. AIoT, also known as "AI+IoT", refers to the integration of artificial intelligence technology and the Internet of things in practical applications. At present, more and more industries and applications combine AI and IoT. AIoT has become the best channel for the intelligent upgrading of traditional industries and the inevitable trend of the development of Internet of things in the future.
As one of the most mature technologies in artificial intelligence application, intelligent voice technology is widely used in AIoT field. B-end medical and educational industries; The new capabilities of smart voice technology in smart phones, smart homes and smart wearables on the c-terminal allow us to see the huge changes this new technology has brought to our lives.
According to the 2018-2022 in-depth research and investment prospect forecast report on China's intelligent voice industry released by China investment consulting, intelligent voice application will be the key direction of artificial intelligence development.
At present, the state is accelerating the industrial layout of artificial intelligence and releasing a number of favorable policies to promote the development of artificial intelligence industry. The advent of 5G era has cultivated a fertile soil for the development of artificial intelligence. We can foresee that intelligent voice, as the new entrance of the next generation of human-computer interaction, will take the lead in taking root in this fertile soil of artificial intelligence, and promote the growth of the whole industry.
Chinese enterprises have shown outstanding performance in the field of intelligent voice, and become increasingly active in innovation and entrepreneurship. A number of leading enterprises such as sogou intelligent voice have accelerated their growth, gaining wide attention and recognition in the international community.
At present, intelligent speech has made a number of major breakthroughs in key technologies such as speech picking, speech analysis and semantic understanding. Intelligent speech technology has presented an overall solution.
Based on the curiosity of the development status of intelligent voice technology in the artificial intelligence industry, a number of well-known domestic technology media came to China's leading consumer chip manufacturer juxin technology together to explore what changes the new era of intelligent voice technology will bring to our lives!
Yongyao Tao , an expert of technology market of torch core, was invited to accept a group interview of the media to solve the mystery of intelligent voice technology and various doubts about the development of AIoT industry.
To help you understand, Yongyao Tao first on the sound of the working principle and intelligent voice technology in the key: sound processing technology to do a simple science. Let's also learn together!
Three main scenarios for sound application:
▪ sound transmission, such as telephone calls, through sound transmission, so that the call function;
▪ sound storage, such as recording, playing CDS and so on;
▪ sound recognition, is the voice interaction experience
The technology of sound pretreatment and sound post-processing is the key to the above scenario.
The post-processing technology of sound is to process the sound after it is stored and played at the same time. Is to audio source such as MP3 and other media decoding playback sound into sound enhancement processing. (for example, virtual bass, surround sound, treble enhancement, EQ, etc.)
Sound pretreatment technology, is the sound is not into the transmission, no storage before the processing. The purpose of sound pretreatment is to make the storage and transmission of sound more efficient and the recognition rate better. (e.g. echo cancellation, noise reduction, sound activity detection, etc.)
Actions in the past 30 years, has been working in the voice processing, such as MP3 recorder (voice of post-processing technology), such as the retention of sound, storage, pretreatment (sound), microphone voice recognition, the echo cancellation and noise reduction of sound (voice before treatment), the late audio enhancement (voice of post-processing technology) and so on.
The key of intelligent voice terminal equipment -- sound preprocessing technology
Sound preprocessing technology is to solve the machine "hear" and "hear" two things, so as to achieve the purpose of "understand".
The "hearing" of sound mainly depends on the microphone. The main form is a single microphone or an array of microphones (multiple microphones are arranged according to certain rules to obtain and process sound in a specific space).
But the speech signal processing algorithm based on the microphone is the key to make the voice "understand". The microphone + algorithm is arranged and combined in different environments to finally reach "hear" and "understand".
We can divide the sound pretreatment into far-field pickup (3 to 5 meters) and near-field pickup (within 1 meter) according to the distance.
Near field pickup, generally using a single microphone, can be in close range, low noise to pick up the voice in line with the needs of speech recognition. But once the smartphone is placed at a noisy distance, recognition rates plumb, and the limitations of a single microphone stand out.
Far field pickup, because of the distance, noise, reverberation and other factors, the far field pickup adopts multi-microphone array, and matches with the far field speech recognition algorithm.
Through Yongyao Tao on the sound working principle and the focus of sound pre-processing technology, sound technology has a further understanding, many media in succession to the torch technology development further questions, we will also be related to the hot questions and answers listed out and share with you!What are the exclusive advantages of Actions in sound pretreatment？
The main advantage of Actions is that we have been doing system integrated SoC, chip hardware technology and software technology can be perfectly combined. With years of experience, the torch core has a good foundation in the accuracy and speed of sound acquisition, and through the sound processing algorithm hardware and other ways, can maintain the chip power consumption and cost performance in the industry leading position.What industries has Actions covered in sound pretreatment technology and what kind of products it has landed？
Our main products now include children's recording pens, story machines, bluetooth speakers and so on. In terms of technology, it mainly includes sound noise reduction, echo cancellation technology, anti-reverberation technology and online speech recognition.
At the same time, torch is expanding into new areas, such as smart homes and smart appliances. In this direction, torch will pay more attention to offline recognition technology (offline voice technology can avoid online technology network instability, difficult networking, privacy exposure, etc.). And in a short reaction time to achieve the control of household appliances. .
The smart ear wearers that torch is currently promoting combine offline and online voice recognition. We can implement some command control through limited command entries, and we can also implement some semantic understanding commands through mobile phone and cloud recognition. In the future, it will be an important direction for the development of the technology to combine with the online identification technology.In the future, what are the difficulties of sound pretreatment technology to be solved？
First, how to make the machine understand is a problem that we need to pay constant attention to in technology. The biggest difficulty in getting the machine to listen is the recognition rate and the false recognition rate, which will be continuously improved in the future. It is necessary to point out that the recognition rate of the machine and the false recognition rate, that is, the probability of hearing right and hearing wrong, have a mutual check and balance relationship.
Second, it is in the environment of picking up sound. Sound pickup is affected not only by its own echo and the reverberation of the environment, but also by all kinds of noises in the environment. All kinds of steady and unsteady noises need to be processed by sound pretreatment technology. How to deal with all kinds of steady and unsteady noises will be a technical difficulty that needs continuous attention in the future.What do you think about the development trend of AIoT and related industry opportunities？
AIoT is an inevitable development trend, and technology is a long-term process of research and development and investment. In the process of technological development, we find that the landing of strong artificial intelligence is a protracted war, but weak artificial intelligence can actually solve some people's needs in some scenarios. Therefore, we also need to seize the market opportunities when some scenario-appropriate AI market opportunities emerge.
AIoT is a protracted war. Short-term and long-term technology research and development and market benefits require our constant balance and development.
To learn more about the interview, please click the interview video below!
As we all know, Actions has been providing professional chips and complete solutions for wireless audio and intelligent ear wear, intelligent multimedia, intelligent computing and Internet of things products for many years. Especially in the field of sound, juxin has launched a number of audio and intelligent robot solutions, in the field of voice interaction experience has been in the industry's leading position, landing products are favored by domestic and foreign brands.
Because Actions in the "sound" field of continuous efforts, more than 20 years of technical accumulation and in-depth cooperation with high-demand brand business practical experience, let the Actions in the industry has been in the leading position.
AIoT, the new era is coming, Actions will continue to pursue the deep development of technology, bring people into a better life experience!