Intelligent speech chip,working principle of off-line speech recognition chip

With the rapid iteration of intelligent hardware,intelligent voice chip,as the core component of human-computer interaction,is widely used in various intelligent terminal devices.The AT700 series speech recognition IC introduced in this paper is a high-performance intelligent AI real-time cognitive execution SoC chip,which focuses on the core advantages of ultra-low power consumption,ultra-high computing power and ultra-low delay,supports off-line local recognition,does not need to rely on the network,and adapts to the needs of multi-scene intelligent interaction.It is a high-quality off-line speech recognition chip solution.
 

AT700 series off-line speech recognition chips rely on professional signal processing mechanism to realize speech interaction function.Through high-precision speech signal sampling technology,the collected analog speech signals are converted into digital signals and stored in the ROM storage unit of chip IC.When the intelligent voice chip device is running,the circuit system retrieves the digital signal in ROM and converts it into a playable voice signal,thus completing the complete voice recognition and broadcasting process.The chip is equipped with three mainstream output modes to meet different hardware development requirements,namely PWM,DAC and ADC outputs,which can flexibly match the circuit architecture of various terminal devices.
 

The PWM output of intelligent voice chip is a common mode in the industry,which is suitable for most conventional voice chip architectures.It has stable structure and strong compatibility,but it does not support continuous volume adjustment and cannot be connected to ordinary power amplifier equipment.The output of DAC is optimized by EQ amplification inside the chip,which supports continuous volume adjustment,external power amplifier equipment and digital precise regulation.The ADC output integrates multiple processing processes such as voice sampling,EQ amplification and signal compression to ensure the integrity and clarity of voice signal acquisition.
 

 

AT700 series speech recognition IC has high integration and comprehensive performance,which is suitable for the development scene of edge intelligent devices.Intelligent voice chip has built-in professional noise reduction module,which can effectively filter environmental noise,and can still accurately capture Chinese and English mixed voice commands in noisy scenes,greatly improving the accuracy of human-computer interaction.The standby monitoring current of digital microphone is less than 100uA,and the standby current of analog microphone is as low as 150uA,which can save energy and reduce consumption for a long time.In hardware configuration,the intelligent voice chip integrates a complete set of power management modules such as DCDC,LDO and POR to ensure stable and efficient power supply for equipment.At the same time,it supports external SPINor Flash,and can store programs,neural network models and user data,which is very extensible.
 

Compared with ordinary voice chips,AT700 series intelligent voice chips are equipped with multiple AI functions,such as visual AI ISP,NLP natural language processing,large model operation,image enhancement,dark light and full color,to realize multi-modal integration of voice and vision.Intelligent voice chip products are widely suitable for high-end intelligent terminals such as edge-side servers,smart cerebellum with body,VR glasses,360-degree BEV scanning devices,AI PC,etc.It is the preferred offline voice recognition chip for industrial and consumer intelligent hardware.
 

As a chip supplier with more than ten years of technology accumulation,RAMSUN provides a series of intelligent voice chip products with wide voltage and low power consumption,and can assist in completing engineering links such as timing matching,PCB layout and firmware adaptation.For detailed data sheets or application notes,please contact RAMSUN.