The VB series module is an AI offline speech recognition product developed by Ai-Thinker. The main chip is the speech brain VB590 launched by Huazhen Company, which has the characteristics of ultra-low cost, high reliability and strong versatility. In the speech recognition technology, it has achieved high-reliable wake-up recognition rate, longer-distance wake-up, lower false wake-up rate, stronger anti-noise ability, faster response recognition time, and pure offline recognition without networking.

The VB series module adopts a high-performance 32-bit processor, the chip has built-in DSP instruction enhancement unit and MCA algorithm hardware accelerator required for speech recognition neural network calculations, AI algorithm and chip architecture are deeply integrated. It equipped with speech recognition algorithm, speech enhancement, reduction Acoustic front-end processing algorithms such as noise provide smart devices with excellent voice control and voice interaction capabilities in the far-field environment. The main chip has been fully and deeply optimized in AI computing power, storage performance, and integration, providing developers with a truly low-cost complete voice AI solution and enhancing differentiated competitiveness.

VB series modules have abundant peripheral interfaces, including UART/I2C/PWM, and simple and friendly secondary development tools, which are convenient for customers to realize single-module voice control application scenarios.

　　Characteristics

• Noise suppression: steady-state, dynamic noise, environmental noise suppression>20dB
• Local speech recognition (based on the latest TDNN deep neural network recognition algorithm), support pure offline recognition
• The comprehensive recognition rate can reach more than 98%
• Recognition time is less than 100ms
• Very low misjudgment rate
• Up to 50 custom entries

　　Core Introduction of Module

• Integrated TG head 32-bit Xuantie 804 processor frequency up to 240MHz
• Adopt 3-4 stage variable length pipeline, equipped with DSP calculation engine and MCA algorithm hardware accelerator supporting NN
• Support 32 interrupt nesting, each interrupt has corresponding priority
• Support JTAG debugging and hardware breakpoint
• Integrated 128KB SRAM 8KB ROM for Boot
• Built-in 2MB SPI FLASH
• Internally integrated audio AD/DA, which can be connected to analog PA and MIC
• Internal integrated LDO, 3.3V input, default output 1.2V
• Built-in watchdog monitor WDT
• Timing and control
• 4 Timers, support a variety of clock source options, 2 of which can be clocked in standby mode and wake up with low power consumption
• 1 RTC, which can be used for accurate timing and low-power wake-up
• Peripherals: 7 channel PWM, UART*2 support high-speed flow control, 10 general-purpose input and output interfaces, each pin can issue an interrupt by rising/falling edge or level detection
• Peripherals: Support I2C, SPDIF and other audio interfaces

VB-01 specification： Chinese English

VB-02 specification： Chinese English

VB-01-Kit specification： Chinese English

VB-02-Kit specification： Chinese English

VB-01 offline voice module access to smart curtain application demonstration：https://www.linkedin.com/feed/update/urn:li:activity:6824259589312249856

VB-01/02 Factory firmware secondary development SDK source code： download

The English version documents is being added, please wait

Shenzhen Ai-Thinker Technology Co., Ltd.
Support: support@aithinker.com
Business: overseas@aithinker.com
Tel: +86-15989589908 (same number on WeChat)
Address: Room 410, Building C, Huafeng Smart Innovation Port, Gushu, Xixiang, Baoan District, Shenzhen