On September 1, iFLYTEK of HKUST held a new consumer product launch conference with the theme of "Sound is the King" in Beijing. Hu Yu, the rotating president of iFlytek, and Li Chuangang, vice president of iFlytek, attended the conference and were on site. Four new products, iFLYBUDS, iFLYBUDS, iFLYBUDS, iFLYBUDS, the first smart TWS headset, were released.
Hu Yu, the rotating president of iFlytek, said that iFlytek adheres to the "To B+To C" two-wheel drive strategy to allow AI applications to serve the public. Adhering to the concept of helping users to work efficiently and live smartly, iFLYTEK’s consumer business has maintained a high growth rate of over 30% for three consecutive years. With the continuous improvement of iFLYTEK’s intelligent hardware product system, C-end revenue accounted for 37.59% in the first half of 2020. On the AI transfer track, we hope that iFLYTEK’s consumer products will not only empower text workers’ recording scenarios, but also become a knowledge management tool for everyone, opening a new era of efficient recording.
At present, the newly released iFLYTEK smart voice recorders SR702, SR101, iFLYBUDS smart headsets have been pre-sold at 0:00 on September 1, and will be available for simultaneous spot sales on Tmall Jingdong and other e-commerce platforms at 0:00 on September 9. The SR901 will start pre-sale in late September. During the pre-sale period of new products, you can enjoy multiple benefits such as installment interest-free plans, pre-determined deposits, and gifts. Participating in the pre-sale has a chance to win the iFLYTEK smart voice recorder SR301.
IFLYTEK’s full range of smart voice recorders has been completed and continues to lead the AI voice recorder market
IFLYTEK released three new smart voice recorders this time, complementing the full range of smart voice recorders, covering the full range of models from entry, flagship, and premium versions, providing consumers with different needs with the most satisfactory choice.
The iFLYTEK smart voice recorder SR101 is an entry-level new product for newcomers and students in the workplace. It combines portable experience, high cost performance and smart experience. It has three color options, starry sky gray, deep sea blue, and ivory white. The standard 1.4-inch touch screen makes the transcription visible in real time. It is also equipped with a combination of 1 directional microphone + 2 matrix microphones, quad-core processing chip, 8GB storage space and 1500mAh battery capacity, no compromise on hardware configuration. SR101 is priced at 599 yuan, free of charge for life and enjoy 3 years of 5GB cloud space service.
【Xunfei Smart Voice Recorder SR101】
The iFLYTEK smart voice recorder SR702 is iFLYTEK’s flagship after the SR701. SR702 adopts starry sky gray color, integrated design of 3D ceramic backplane and aviation-grade aluminum alloy. It uses high-density 2600mAh high-capacity battery, equipped with 2 directional microphones and 6 matrix microphones, equipped with 3.5-inch HD screen, and equipped with 8 million pixels Digital zoom rear camera. The iFLYTEK Smart Voice Recorder SR901 Premium Edition uses a piano black full ceramic body, equipped with a 6.01-inch AMOLED screen, 64GB storage space and 4000mAh supporting 18W fast charge large-capacity polymer lithium battery.
【Xunfei Smart Voice Recorder SR702】
Xunfei smart voice recorder SR901 has a luxurious three-camera configuration and 2 Harman MEMS directional microphones + 10 omnidirectional microphones. Through the fully upgraded transcription engine, SR702 and SR901 are equipped with the industry's first offline transcription, OCR graphic recognition, video subtitles and other new functions, which once again broaden the ease of use and usage scenarios of transcription records. The SR702 is priced at 3699 yuan, and the SR901 is priced at 5999 yuan. The purchase of two products comes with free lifetime membership services, including transcription and translation services. At the same time, from the day of activation, an additional 3 years of 20GB cloud space service will be given.
【Xunfei Smart Voice Recorder SR901】
The above three new products all follow the four major standards of iFLYTEK’s smart voice recorder-accurate recognition, remote recording, clear recording, and full language. In terms of recognition accuracy, the new voice recognition module of iFLYTEK’s smart voice recorder adopts a voice recognition framework based on a deep full-sequence convolutional neural network, with a recognition accuracy of 98%; in terms of recording distance, the new product of iFLYTEK smart voice recorder has the longest pickup distance It can be up to 15 meters; in terms of recording clarity, the new iFLYTEK smart voice recorder uses AI algorithm to reduce noise to process the front-end voice signal to ensure the accuracy of pickup, and provide high-definition lossless sound quality audio format; all languages, iFLYTEK Smart The new recorder also supports Cantonese, Henan, Sichuan, Chongqing, Yunnan, Guizhou, Tianjin, Hebei, Northeastern, Gansu, Shandong, Taiyuan, 12 dialects, Tibetan, Uyghur 2 ethnic minority languages transcribed, Chinese, English, Japanese, Korean, Russian, French, Vietnamese, Spanish 8 languages transcribed and IT technology, finance and trade, culture and sports, medical care, politics and law, operators, education and 7 major fields Transliteration optimization. At the same time, the Chinese and English mixed reading recognition function can also automatically recognize the voice content mixed with Chinese and English when transcribing.
Thanks to the newly upgraded transcription engine, iFLYTEK’s SR901 and SR702 smart recorders are the first in the industry to be equipped with offline transcription functions. Without 4G/WIFI network, users can also realize real-time voice conversion to text, and the text content after the transcription is saved immediately. The offline transcription accuracy rate is as high as 95%, and there is no fear of network disconnection. At the same time, offline transcription also prevents the leakage of highly confidential information, ensuring user information security and the realization of unique transcription requirements. At the same time, through the SR901 and SR702 cameras, the text content on the picture can be extracted and presented in real time, which is convenient for users to realize the mixed recording of important audio and graphic information for later recording and sorting. At the same time, these two new products also support video recording, real-time radio and real-time generation of subtitles. At the same time, subtitles can also be edited later to create a new AI interactive experience for recording important information in multiple scenes.
In the front-end noise reduction processing of Xunfei smart recorder, through Xunfei's original neural network and traditional signal processing deep combination of noise reduction algorithm (SSA-IME), it can effectively reduce environmental interference noise, and effectively eliminate the voice that interferes with the speaker. This greatly reduces the processing difficulty of speech recognition. In the case of no speakers, the distance of 3-10 meters is increased by 30%, and in the case of speakers, the effect of super long-distance (20-50m) speech scenes can reach more than 90%. In order to meet the needs of users for quick drafting, iFLYTEK has invested heavily in speech recognition post-processing technology and applied it to iFLYTEK's smart recording pens, and successively launched text comprehension capabilities such as tone filtering, spoken language regularization, text semantic segmentation, and text intelligent summarization. For the multi-person conversation conference scene, the speaker separation scheme of separation and ASR joint modeling is launched. The accuracy of speaker separation reaches 95%, allowing users to quickly retrieve target speaker data and improve editing experience.
IFLYTEK’s first TWS smart headset released, real-time voice transcription creates a new standard for mobile commerce
Also released at the same time as the three new products of Xunfei smart recorder is Xunfei's first TWS smart headset iFLYBUDS. With the three core AI functions of real-time call transfer, intelligent dial recognition, and call translation comparison, it can easily meet the needs of business people to free their hands, record important information and communicate calmly in multiple scenarios such as self-driving, commuting, and meetings.
IFLYBUDS uses pearl white color matching and arc-shaped aesthetic design. The earphone weighs only 4.6g and has a semi-in-ear bionic design that is comfortable to wear for a long time. The device is equipped with a flagship chip to ensure connection stability and ultra-low latency. In addition, iFLYBUDS also has pleasing sound quality. Through intelligent noise reduction, it can also bring a clear and smooth call experience in noisy environments such as airports and gyms. iFLYBUDS supports intelligent voice assistant, just say "Xiaofei Xiaofei" to wake up the voice assistant quickly. At the same time, the features of iFLYBUDS also include: support for tapping headset interaction, and support for custom functions; wear detection, pause playback when you take off the headset, put on the music again to continue. Not only that, the iFLYBUDS battery life of iFLYTEK smart headset can satisfy daily business communication.
One of the core functions of iFLYBUDS, iFLYBUDS, supports regular calls and VoIP (WeChat voice, Tencent meeting, DingTalk voice, Skype, Zoom, etc.) in real-time call transfer, and you can answer VoIP conferences anytime, anywhere. And for multi-person conferences, it also supports automatically distinguishing the speakers, making it more convenient to find the content of the call. One-click recording at the beginning of a call, and the call is converted into text, and the transcription accuracy rate is above 98%. After the call is over, an intelligent summary can be formed to automatically refine the key content, and it also supports quick search, multi-terminal sharing (mobile and computer) and editing of the content of the call.
Intelligent dial recognition allows iFLYBUDS to solve the problem of numerous contacts in the address book of business people, and the pain points of multiple names and homophonic contacts. In the case of duplicate names and homophonic dialers, iFLYBUDS can be distinguished in many ways. For example, Zhang Ziyi's chapter and Lizao chapter; distinguished by the name of the place of attribution, such as Mr. Li of Beijing and Mr. Li of iFLYTEK. Avoid complex search, and double the dialing efficiency. Compared with voice assistants such as Siri, the voice dialing function of iFLYBUDS focuses on dialing, smarter and more accurate, and has obvious advantages: it is more natural to speak, you can dial by just saying the name; pure voice interaction, no need to look at the screen; one sentence is done, avoiding repetition .
In the foreign language telephone communication scene, facing the rare professional vocabulary such as industry terminology, slang, etc., language communication barriers are often encountered. In response to this, the third core function of iFLYBUDS for business people is to compare the translation of the call. It can translate the content of the call into Chinese characters in real time during a call in English, assisting in English communication. It is the assistant translator on your ears, making communication more confident and simple. More languages will be supported in the future.
Twenty-one years of deep accumulation of AI technology, iFLYTEK’s new consumer products let consumers buy with confidence
At the press conference, Li Chuangang, vice president of iFLYTEK, stated that “the two-wheel drive of technological innovation and consumption upgrade will present huge opportunities for disruptive innovation and give birth to new leading brands. iFLYTEK’s smart voice recorder and iFLYTEK’s smart headset are both. The practice of a concept. Through the trinity of top AI voice technology, high-quality hardware and software, to create an extremely convenient experience, help users effectively improve efficiency and enjoy high-quality intelligent life.
In addition to excellent market performance, iFLYTEK's consumer products have also been highly recognized in the industry, which is inseparable from the 21 years of intensive cultivation of iFlytek in the field of intelligent voice. Take the speech recognition technology applied to the speech transcription function as an example. As early as September 2016, in the industry’s most influential, highest-level international multi-channel speech separation and recognition competition CHiME-4, iFLYTEK Won the championships of all three events; in 2018, we once again swept the championships of all four events in the CHiME-5 competition and greatly refreshed the best results in each category; CHiME-6 in 2020 is known as "the hardest in history "Speech Recognition Task", iFLYTEK not only won the two competition tasks of multi-channel speech recognition with a given speaker boundary, but also refreshed its 46.1% speech recognition error rate champion in CHiME-5, which dropped to 30.5%.
In terms of speaker recognition technology, in the recent limited training data task of the VoxSRC International Speaker Recognition Competition (opening the ranking stage), HKUST IFLYTEK broke the world record, and the EER was reduced to 0.81%. At the same time, iFlytek obtained an EER of 0.63% on the public test set VoxCeleb1 in academia, which is the state of the art (best result) so far. The above two new records show that iFLYTEK’s speaker recognition technology stands in the world Forefront.