Research On Active Mechanism Of Low Power Speech Human Machine Interactive System Based On Speaker Features

In the actual voice interactive system mobile devices,Standby times is the major problems plaguing mobile devices power consumption.It's rely on voice human interactive system worktime,System power consumption comes from the power consumption of the system's core mircroprocessor microcontroller.Therefore,how to control the working status of the microprocessor is very important.Based on source analysis of HMI's power consumption,this theis presents a dual-core 8bit RISC microcontroller and 32 bit RISC microcontroller voice interactive system wake-up mechanism,The first wake-up mechanism through the voice amplitude comparison circuit real-time comparison of the general environmental fluctuations wake-up 8bit RISC microcontroller.The second wake-up mechanism through 8bit RISC microcontroller text-related speaker identity verification wake-up 32 bit RISC microcontroller.Based on analyzing the importance of text-dependent speaker technology to voice HMI system,This theis presents a method of intelligent comprehensive screening for text-dependent speaker recognition,Namely speech text recognition and speaker recognition two recognition systems,Then through the way of intelligent judgments for identity verification,Speech text recognition using mature technology platform.The two recognition rates do not affect each other.This theis propose a weighted principal component algorithm with limited instruction length,Improve Speaker Recognition Rate and Increase Text-dependent Speaker Identity Confirmation Wake-up System 32-bit RISC Microcontroller Wake-up Rate.This theis experimentally studied the relationship between recognition performance and instruction length and choose a valid instruction length,Both to improve system security and reduce system power consumption,It provided the basis for the instruction of voice HMI system.Through experimental and analysis results show,Instruction Text The text-dependent speaker recognition rate increased by about 2.86% over the 2-7s length.Dual-core microcontroller at the same time working mode and only increase in power consumption is about 8.8%;At 8bit RISC microcontroller work,32 bit RISC microcontroller is not working situation,The power consumption reduction rate is about 68.2%;Dual-core microcontroller shutdown mode,the power reduction rate is about 99.03%;Dual-core MCU1,MCU2 shutdown mode Compared with the normal single-core MCU shutdown mode power consumption decreased by 20.38%.