Introduction to Chipintelli¶
Company Profile¶
Chipintelli was established in 2015, with its headquarters in Chengdu. Focusing on the development of speech AI chips and AI algorithms, the company is a full stack technology enterprise in the industry that has the capabilities of AI chips, AI algorithms, development platforms, and software and hardware solutions at the same time. Now, it has grown into a leader in the field of artificial intelligence speech chips, and leads the industry in the market share of end-to-end speech chips.
ASR chip¶
Since its establishment, Chipintelli has launched a number of chips, all of which have the advantages of high performance, high integration, high reliability, low power consumption, low cost, etc
In 2024, Chipintelli launched two major AI voice chip series: the CI13LC series (including CI13081, CI13082, CI13161, CI13162, CI13241, CI13242, CI13322, as well as CI13161P and CI13162P) and the CI23LC series (including CI23161, CI23162, CI23241, CI23242). Both series are equipped with the new-generation self-developed brain neural network processor BNPU V3.5. The system main frequency is as high as 210MHz, and it has a built-in 288KByte SRAM. It integrates a PMU power management unit, an RC oscillator, a single-channel high-performance and low-power Audio Codec, as well as multiple peripheral control interfaces such as UART, IIC, PWM, and GPIO, and fully supports the offline voice recognition function. Among them, the CI13161 and CI13162 chips of the CI13LC series are fully compatible with the third-generation chips CI1311 and CI1312, and there is no need to change the peripheral hardware. The CI13161P and CI13162P further integrate a high-performance power amplifier unit, which further enhances the audio playback ability. On this basis, the CI23LC series integrates a high-performance and low-power Bluetooth BLE. With only a small number of peripheral devices, it can achieve high-performance voice recognition, voice noise reduction, and strong environmental noise suppression capabilities. Among them, the CI23162 and CI23242 also support Bluetooth BLE5.4 sending and receiving, and the device can be remotely controlled through a mobile phone applet. With both excellent performance and high cost-effectiveness, it provides a comprehensive solution for intelligent voice products.
In 2023, the AI voice BLE chip CI231 series (CI2311, CI2312) will be launched, which integrates BNPU V3 and has a system main frequency of up to 220MHz. It is equipped with up to 640KByte of SRAM, which can provide more sufficient computing power to the main control application solution. Adopting TSSOP24 packaging, the pins are wider, making it easier to produce SMT chips with lower SMT costs. In addition, an RC oscillator is integrated, which adds serial port adaptive function and eliminates the need for crystal oscillator applications. It can achieve functions such as speech recognition, voiceprint recognition, command word self-learning, speech detection, and deep learning noise reduction, and has strong environmental noise suppression ability.
In 2022, the AIoT chip CI230 series (CI2305, CI2306) will be launched. This series of chips integrates the brain neural network processor BNPU V3 developed by Chipintelli, as well as WIFI and BLE functions. It is a three in one AIoT chip that integrates “voice+WIFI+BLE”, and can achieve various intelligent voice offline product hardware solutions with only a small amount of peripheral devices such as resistors and capacitors, making it highly cost-effective. In terms of speech algorithms, it supports speech recognition, single microphone deep learning noise reduction, offline natural language processing, dual microphone deep human voice separation, command word self-learning version 2.0, and can support up to 500 command words. CI2306 also supports voiceprint recognition and offline NLP.
In 2021, the CI13 series of voice AI chips (CI1301, CI1302, CI1303, CI1306) will be launched. This series of chips integrates BNPU V3, with a high main frequency (240MHz), larger S RAM, and stronger computing power. Adopting two types of packaging, SSOP24/QFN40, it can adapt to different applications and SMT requirements. Higher integration, streamlined periphery, except for MIC and speakers, board level only requires resistance capacitance and PA chips. It can achieve functions such as single microphone noise reduction, AEC, dual microphone enhancement, and dual microphone orientation. In addition, CI1303 and CI1306 also support voiceprint recognition, 500+command entries, and offline NLP.
In 2020, the voice AI chip CI112 series (CI1122) was launched, which integrates BNPU V2, with lower cost and higher cost-effectiveness.
In 2019, the CI110 series of noise reduction and recognition integrated speech AI chips (CI1102, CI1103) were launched, which support offline far-field speech recognition, offline+online speech recognition, and offline+AIoT speech scene control. The integration is more comprehensive (integrating noise reduction and pickup, microphone array processing, ASR, MCU control, and other functions), and the cost is lower (at the same level as DSP and MCU chips), Lower power consumption (standby power consumption is less than one tenth of that of AP chips that achieve the same functional performance). After evaluation of scientific and technological achievements, the level of achievement has reached international advanced level (certificate number: 202051ZK3713), and has been selected for the “AI Chip Technology Selection Catalogue 2020” initiated by AIIA.
In 2018, the AI voice chip CI1002 was launched, which maintains the performance advantage of CI1006 at a lower cost and has a wider range of applications.
In 2017, Chipintelli launched the first deep neural network AI voice chip CI1006 in China, which enabled intelligent voice offline control of smart home appliances and established Chipintelli’s leading position in the field of offline voice home appliances. According to the evaluation of scientific and technological achievements, it is the first in China (certificate number: 9562018Y0009) and the only voice chip selected for the book “The Wave of Artificial Intelligence”.
AI algorithm¶
As a “chip+algorithm” company, Chipintelli has a full range of voice interaction technologies with completely independent intellectual property rights, including single and microphone array voice enhancement, noise suppression, echo cancellation, low-power wake-up, local semantic understanding, voiceprint recognition and other voice algorithms. It also introduces local self-learning technology, allowing users to customize entries, solving the problems of dialect recognition and user pursuit of personalization. The highest speech recognition rate can reach over 97%, and the recognition distance can reach up to 10 meters, which is at the leading level in the industry.
Solution and application¶
Based on the company’s self-developed speech AI chip and AI algorithm, Chipintelli has launched a wide range of product solutions, which can be used in smart home, smart home appliances, smart lighting, smart toys, smart robots, smart cars and other fields. And open ☞Chipintelli Speech AI Development Platform, designed to help developers develop more efficient and convenient intelligent speech products.
After years of accumulation, the company currently leads the industry in the number of product solutions based on end-to-end speech chips, and hundreds of partners have developed thousands of product projects based on the company’s chip products. The company will continue to develop AI speech chip products and application solutions that are more suitable for the market and customer needs, aiming to accelerate the market penetration of intelligent speech products through continuous iteration of technology and products, and make people’s lives better with intelligent speech.
Company Mission¶
Lighting up life with AI