[Open Source Project] Voice Recognition Controlled AI Night Light Based on ESP32

1. Project Introduction Using ESP32-S3N8R8 module as the main control chip, the S3 core adds instructions for accelerating neural network calculation and signal processing, which allows us to use it to quickly parse the trained voice model for voice recognition. 2. Principle Analysis This project consists of four parts, power supply part, LED lighting part, main control part, and voice recognition expansion part. This project mainly receives and processes voice signals through a microphone, extracts human voice for analysis and comparison, and performs corresponding control operations when the sound meets the instructions. 2.1 Power supply circuit The TYPE-C-16P interface is used as the power supply interface, and the corresponding USB data pin is used to connect to the corresponding USB pin of S3 (USBD+ IO20), (USBD-IO19), and USB is used directly for downloading and debugging without converting to serial port signals. Add 5.1K pull-down resistors to the CC1 and CC2 pins to facilitate recognition and configuration of different hosts. Using AMS1117 as a 5V to 3.3V step-down LDO, ESP32S3 consumes a large current when turning on wireless radio frequency or performing voice analysis operations. In addition, there are other peripheral circuits. When selecting the power chip, the output current is at least 600mA. The output current of AMS1117 is 1A, which can meet the requirements. 2.2 The LED lighting circuit uses four RGB three-color lights distributed around the board. Different display effects can be obtained by changing the brightness of different colors of the RGB three-color lights. Considering that the red, blue, and green lights are inconsistent in the required current, different resistors are used to be connected in series in the corresponding branches. The brightness can be unified by adjusting the resistance value later. At the same time, the lights of each color are strung together, and they are uniformly turned on and off through the SI2302N channel MOS tube. The brightness can also be adjusted by controlling the on time through PWM. 2.3 The main control part uses the ESP32S3N8R8 module as the main control chip. It should be noted that if voice recognition is required, the required resource library is relatively large. It is recommended to select Flash and PSRAM above 8M. Note that in modules with OSPIPSRAM (i.e., the built-in chip is ESP32-S3R8 and above), pins IO35, IO36, and IO37 are used to connect to the OSPIPSRAM integrated inside the module and cannot be used for other functions. Here we need to mark the corresponding pins with non-connection marks. In ESP32S3, there are a total of 4 Strapping pins. When assigning pins, try not to add pull-up or pull-down resistors to these pins to change their default state. 2.4 The voice recognition expansion part uses an I2S digital silicon microphone to receive voice signals. At the I2S signal line, in order to obtain a better anti-interference effect, you can try to connect a small resistor in series for impedance matching. Use the D-class power amplifier chip of the I2S signal for voice output. In the sound output path, add magnetic beads and capacitors to form an LC filter circuit to reduce output interference. The external speaker is connected using the GH1.25*2P interface. ESP32S3 has two I2S controllers, and all IOs can be reused as I2S pins through the internal matrix, which can be assigned at will. Adding a vibration sensor, when the hand hits the desktop, different lights can also be switched, increasing playability. The vibration sensor has a spring structure inside, which can be simply understood as a spring button.