AI voice chips pile up and emerge, and emerging markets rise quietly

Friends who pay attention to speech recognition will certainly notice a phenomenon. Recently, various companies have successively launched AI speech dedicated chips, such as yunzhisheng and going out in May. Not long ago, rokid and spitch, the four AI speech enterprises collectively bet on AI speech chips. What are the market opportunities behind them?
Promotion of intelligent speaker
Intelligent speaker is a hot topic frequently. It is predicted that with the competition between Amazon and Google in the market, the total shipment of global smart speakers may exceed 60 million units in 2018, while the domestic market scale is expected to grow at least 10 times, and the market scale will exceed 10 million.
Behind the strong outbreak of the domestic smart speaker market, it also drives the competition for voice chips. Especially when the chip scheme in the smart speaker starts to change from general chips to special voice chips, the competition in this part of the market is more intense. For example, Alibaba's tmall elf uses MediaTek mt8516 voice chip, while Xiaomi Xiaoai speaker uses Jingchen A113 chip. Better experience and lower price are the basis of speaker business war.
In addition to the common smart speakers, many hardware devices in home and office scenes have also begun to be voice intelligent, which has exacerbated the outbreak of the market of AI dedicated voice chips. Therefore, in this situation, many voice technology companies began to layout a large number of voice chips and voice interaction solutions with the accumulation of technologies such as voice recognition, natural language processing and voice interaction design.
Why chips?
When it comes to the competition of speech recognition, perhaps the first reaction of many people is the competition of artificial intelligence algorithms. But the reality is that it is difficult for startups to do pure algorithms for a long time. The algorithms have bottlenecks and ceilings, so they need to be combined with hardware to realize commercial value.
In contrast, in the field of speech recognition, this field is very specific and needs to realize the corresponding functions with hardware with high cost performance and the best performance power consumption ratio. However, as an emerging market field, when the equivalent has not accumulated to a certain extent, the traditional chip giants are unlikely to vigorously enter the market with tens of millions of shipments. After all, they pay more attention to the market. Therefore, there is a certain market vacuum. Startups focusing on algorithms need to ensure their core values and interests in the selection of chips, and can develop the most suitable chips based on their own algorithms and capabilities, which has become an inevitable way for many algorithm companies.
Therefore, we can see that enterprises that once focused on artificial intelligence algorithms have tested AI voice chips, as evidenced by several domestic voice technology start-ups successively launching AI voice dedicated chips. To borrow the words of the industry, it is "there is no doubt that voice companies will die if they do not make chips", and rokid CEO Zhu Mingming also said that "there is no doubt that voice companies do chips".
Exclusive chip
Of course, the chips launched by these enterprises are not general chips, but AI voice chips more suitable for their own artificial intelligence algorithms, which not only limits the functions, but also limits the scope of enterprise cooperation to a certain extent.
Compared with those general-purpose chips, voice chips have outstanding characteristics. Considering both computing power and low power consumption, they use the CPU most suitable for voice processing; Highly integrated voice SOC, support multi-channel microphone array interface, integrate codec module / DSP module, WiFi / Bluetooth module, etc; Support echo cancellation, noise suppression, sound source location, speech enhancement and other technologies in speech algorithm, and have good sound value adjustment function; Moreover, most of the application scenarios of voice chips are intelligent terminals, which can integrate neural network units and localize some intelligence trained in the cloud.
The history of exclusive voice chip can be traced back to July 2013. At that time, Changhong voice chip integrated various speech enhancement functions on the basis of speech recognition. But it was not until 2015 that voice chips began to rise rapidly. The development of voice chips is inextricably related to smart speakers and smart homes. Many voice chips are even specially built for this purpose.
But interestingly, except MediaTek, other voice chips are basically from small and medium-sized chip companies, while chip giants such as Qualcomm and Intel have not launched voice exclusive chips. On the one hand, compared with mature markets such as mobile phones and computers, the market scale of voice chips is very small, which is not enough to attract the attention of chip giants. At the same time, it also reflects that they have made slow progress in the layout of voice chips.
Chuang Weixin has something to say
Nowadays, the application scenarios of speech recognition are more and more extensive. The application of intelligent speech technology continues to expand in all fields, from smart home to automobile intelligent mobile devices, from unmanned bank to new retail, and then to security. At the technical level, artificial intelligence algorithms encounter bottlenecks. It is difficult to differentiate products in the industry. At the same time, the space for performance improvement is realized by the exclusive AI voice chip.
In this emerging market segment, MediaTek is unlikely to achieve monopoly, and other chip giants have not yet entered the board, which provides an ideal development space for many AI companies to transform into chips.
However, from the perspective of the audience, the integration of AI in chips and terminal devices can effectively improve the quality of life. Both the current smart speakers and the unpopular VR glasses will promote the transformation of our lives.

Commax-Tech Electronic Co., Ltd      Electronic component specialist

B23, second floor, ASEAN building, Longhua District, Shenzhen

sales@commax-tech.com

https://commax-tech.com

Keyword：AI Voice Chip   Image Recognition   Start-up Chip Company   Chip Downstream Algorithm   Chip Application Solution   Smart Speaker   Artificial Intelligence   Commax-Tech Electronic