Contact us
Home › Product CenterBluetoothVoice recognitionAISTARTEK
The voice recognition chip is also called the voice recognition IC. Compared with the traditional voice chip, the biggest feature of the voice recognition chip is that it can recognize the voice. It allows the machine to understand the human voice, and can perform various actions according to the command, such as blinking the eyes. , Move your mouth (smart doll). In addition, the voice recognition chip also has high-quality, high-compression rate recording and playback functions, which can realize man-machine dialogue. The speech recognition chip has the characteristics of ultra-low cost, high reliability, and strong versatility. In terms of voice technology, it has realized high-reliable wake-up recognition rate, longer-distance wake-up, lower false wake-up rate, richer number of voice control commands, stronger anti-noise ability, faster response and recognition time, and no networking Pure offline recognition. 
Aistartek Electronic Technology Co., Ltd. subversively combines AI and analog computing chip technology to create a new deep neural network processor with ultra-low power consumption and high performance, which provides unlimited possibilities for edge computing functions such as voice wake-up / recognition and image classification / recognition to settle in mobile terminals.

AT68xx is an off-line intelligent speech recognition chip with ultra-low power consumption, integrating advanced voice activity monitoring (VAD), voice print recognition, automatic noise elimination neural network, efficient power management module, digital and analog voice signal input interface, ARM Cortex-M0 core, and equipped with GPIO, UART, SPI, I2C, I2S and other on chip and off chip devices. In Bluetooth headset, smart remote control, smart home and other scenarios, AT680x can realize offline voice recognition with smaller size and extremely low power consumption.

Features and advantages
•CPU and storage -Up to 2 channels of I2S: 1 channel of voice output port, 1 channel of voice input port
-1 ARM Cortex M0 core •Power management
-64 KB SRAM: 32KB IRAM and 32KB DRAM -Single power supply can optimize the power consumption of integrated 
Support external SPINOR Flash for storing applications, system and reduce peripheral components
neural network models and user data -Internal integration of DCDC, LDO, POR and other power management modules 
•Low power consumption: standby monitoring current is less than 100uA  to ensure stable and efficient power supply system
(digital microphone)/150uA (analog microphone) •Clock management
•Dedicated neural network -Support 32.768 kHz crystal oscillator, no crystal, numerical control oscillation, 
-low power consumption external input and other clock supply modes
-Near memory -Internally integrated numerical control oscillator (DCO and LPO) and digital frequency 
-Automatic noise elimination lock loop (ADPLL), supporting intermittent clock calibration function
-Speech recognition •System peripherals
-Support offline voiceprint recording and recognition -At most two SPI (one master and one slave), SPI slave can directly access registers
 -Supports up to 2 wake-up words and 60 command words -Up to two I2C channels (one master and one slave),
•Audio interface supporting external I2C sensors and direct access to registers
-2-way 14 bit audio ADC, with low-power programmable gain amplifier  -Up to two UARTs can be loaded and upgraded through the serial port
 and low-noise programmable microphone bias power -Up to 2-way timer
-Support voice activity monitoring (VAD), compatible with MEMS microphone  -Up to 4 PWM outputs
and bone voiceprint -Up to 24 multi-functional GPIOs
-Up to 2-way PDM, supporting two-way digital microphone •Support deep sleep mode, which can be awakened by external interrupt

Part Number CPU Digital microphone PDM Timer GPIO Flash Package Download
AT6811ABF1 ARM Cortex-M0 2-way 2 24-way 512KB QFN48 NULL
AT6802ABR1 ARM Cortex-M0 2-way 2 9-way 512KB QFN20 NULL
AT6803ABR1 ARM Cortex-M0 0-way 2 6-way 0 WLCSP-14 NULL