Human-computer interaction new blue sea - speech synthesis

Posted on

Many of today's people can access artificial intelligence products and application scenarios, such as: smart home, smart robots, smart speakers, virtual idols, audio books, education industry and so on. Whether it's at home or in an outing scene, the beautiful synth sound seems to have become a new character IP, although sometimes it's not smart enough, but many electronic enthusiasts are willing to give more patience and gradually turn them into partners in life. .
Behind everyone's enjoyment, it is the practitioners of the industrial chain that promote the subtle advancement, and the accumulation of the moon in the day can be seen in today's results.
How much do you know about the technology chain behind the smart speakers?
First of all, let me give you a general idea of ​​which intelligent modules are used to support the emerging interactive mode. The voice interaction process is divided into three steps: information input, information processing and information output. The corresponding technologies are speech recognition, semantic analysis (natural language processing) and speech synthesis.
Knowledge 1: Voice recognition and semantic analysis are early in the Red Sea
Human-computer interaction new blue sea - speech synthesis
In the past few years, speech recognition has always been a hot spot of social concern. The University of Science and Technology and Sogou have been chasing after the recognition accuracy. The natural language processing technology in semantic analysis has also been applied in large-scale commercial applications. Chatting robots have sprung up, but the voice synthesis technology alone is unattended, and it is particularly unpopular.
However, demand for the speech synthesis market has surged since 2018. According to senior sources: "Many of the big cattle engaged in speech recognition are now transforming into speech synthesis. In some companies, for example, her previous emphasis on synthesis is not particularly high, but since last year, the focus has gradually shifted to synthesis."
Knowledge Point 2: Speech Synthesis New Blue Ocean, have you missed it again?
Speech synthesis, also known as text-to-speech conversion, refers to the technique of producing artificial speech through mechanical and electronic methods. In the past, the voice interaction market environment has been recognized by companies for many years, and there is not much attention to synthesis. However, in many fields, the synthesis effect of sound is getting higher and higher. For example, all kinds of smart toys, home appliances, etc. must have voice interaction.
In short, the previous sounds can be used as long as they are available. Now, with the richness of personalized products, the application scenarios and user experience of the products are better displayed. The demand for customization is getting higher and higher, and the product and the scene have a good fit after the sound comes out.
Throughout the domestic voice synthesis market, giants such as Keda Xunfei, Baidu, and Jietong Huasheng have already launched an open platform for speech synthesis to provide standardized services. Qi Guanqiong believes that standardized services are difficult to meet individual needs, especially for SMEs.
Knowledge point 3: Becker technology because of concentration, so ALL IN
As a midstream enterprise in the voice industry chain, the source of the customer's source can reflect the market environment. According to Qi Guanqiong, there are two types of customers in Baobei Technology: one is a large company, they generally have the ability to do synthetic technology research and development, but they lack data and need to find us to do it; the other is small and medium-sized enterprises, which have been customized for SMEs this year. The service is relatively strong. In addition to lack of data, algorithms have no ability to do it (because the cost of R&D is very high and experienced developers are very difficult to find), Becker Technology provides a systematic solution.
It is worth mentioning that the standard technology not only has synthetic technology, but also a long self-owned voice library. It is understood that in the TTS front-end training set, the standard has more than 150,000 rhythm training sets, more than 150,000 sentence-based part-of-speech training sets, more than 150,000 sentences of multi-song training sets and 100,000 TN training sets.
At present, Becker Technology has established long-term and stable cooperative relationships for Baidu, Tencent, Didi, Sogou, Rokid, Storm Group, Going Out, Roobo, Himalayan FM, Cheetah Mobile and many other customers.
Do you think speech synthesis is as simple as that? Answer: NO
After the completion of the voice data product delivery, the service has only begun, there is a good sword, and it needs a first-class swordsman to use it. After the voice data product is delivered, the customer still needs to do some debugging and improvement, but if the process is not handled well, the product experience will be greatly reduced. The customer also wondered: What went wrong? Marke Technology is also acting as a master to guide the swordsman how to use a good sword.
"Teacher" Biaobei Technology recently released the TTS (a kind of speech synthesis application) evaluation system, which is solving the headache problem for customers.
Generally speaking, the TTS system can be divided into two parts: the front end and the back end. The front end completes the normalization of the input text, word segmentation, pronunciation prediction, prosody structure prediction processing; the back end models the sound, and learns the sound parameter synthesis sound through learning. . Due to the complexity and openness of natural speech itself, the front-end processing part is difficult and needs to be widely covered, which has always been the focus and difficulty in the field of speech synthesis.
Specifically, the standard technology evaluation system is divided into three major modules, objective evaluation, rating evaluation and comprehensive evaluation.
1. The synthetic objective evaluation is mainly embodied in four modules: prosody, polyphonic words, numeric symbols and participles in the front end of the synthetic system;
2. Synthetic score evaluation. This evaluation module scores the voice of the TTS synthesis system in two different ways, horizontally and vertically, with the evaluation personnel representing different TTS user groups. The purpose is to let users understand the optimizable space and market competitiveness of their TTS system;
3. Comprehensive evaluation, through in-depth analysis of 10 samples of synthetic test sets, comprehensive analysis of synthetic systems from text analysis problems, prosodic level prediction problems, acoustic parameter generation problems and vocoder problems, Form an evaluation report. The evaluation report is divided into two parts: the first part is generated by machine synthesis. After the user downloads the test set, it can generate an online test result; the second part is a more in-depth manual evaluation. All are currently free reviews. Through evaluation, users can get a deeper understanding of the key issues of the synthesis system, and achieve the purpose of improving the system synthesis effect more efficiently.
It is estimated that you will probably know that a smart speaker with a smooth dialogue is hard to come by. Although we only introduce the TTS evaluation system of the third part of the standard synthesis of speech synthesis, you can still feel the voice interaction. It is the future, and this kind of future is based on thousands of R&D personnel, engineers, and voice data service providers working on the day and night. They listen to every sentence you say, every voice they have to you. Interactions are all hopeful, and they are fully committed to your quality synthesis every time you synthesize your speech...

16MM Metal Switches

16MM Metal Switches

Yeswitch 16MM Metal Switches could be divided into aluminum casing and stainless steel casing and also could be divided into Momentary Switch and self-locking Metal Push Button Switch .

This 16MM series Waterproof Push Button Switch offer a long life expectancy, could used in Industrial control instruments, Medical equipment, Security monitoring equipment, Vehicle peripherals, Audio-visual equipment and Energy storage equipment,etc.

Metal Push Button Switch


In addition, All casings are made through high-precision lathes, and the polishing and plating are rigorously screened. The metal fittings inside the metal switches are made of brass gold plated material, so the switch can have good conduction function and stable quality after long-term use.

The 16 series illuminated metal switches offer a long life expectancy, water resistance to IP67 ratings, and ring or power symbol illumination.This switch has a 16mm panel cutout size. Additional options include a high, high flat or rounded bezel option, and your choice of a solder lug or wire lead termination.

As for the indicator light , we offer customized service, customers could choose the effect , shape and the color. What is worth mentioning is that our indicator light could offer double color, which could offer our customers more choose in the item of light. Meanwhile, on the item of terminal shape, we could also offer customized service, we have solder terminal and free wire length could choose, 500mm is the normal standard.


Momentary Switch


This serious metal Push Button Switches are of high quality and reliable products. The switch has passed IP67 dust proof and waterproof certification, which indicates it can be operated in harsh environments. Moreover, all materials could meet the European and American environmental protection requirements , for example, UL and ROHS certificate.

16Mm Metal Switches,Metal Push Button Switch Momentary,16Mm Momentary Metal Switch,16Mm Metal Push Button Switch

YESWITCH ELECTRONICS CO., LTD. , https://www.yeswitches.com