Adafruit BrainCraft HAT for Raspberry Pi
Engineer Ladyada has been designing machine learning devices compatible with Raspberry Pi: the BrainCraft HAT and Voice Bonnet!
The purpose of the BrainCraft HAT is to enable you to “create a brain” with microcontrollers and microcomputers for machine learning at the EDGE. In ASK AN ENGINEER, we spoke with Pete Warden, the technical lead from Google Brain TensorFlow Group, about the optimal design of the development board.
BrainCraft HAT
This is what we designed! The BrainCraft HAT features a 240×240 TFT IPS display, a cable slot for connecting imaging cameras, a 5-way joystick for UI input buttons, left and right microphones, stereo headphone output, stereo 1W speaker output, three RGB DotStar LEDs, and a 3-pin STEMMA connector on two PWM pins so they can drive NeoPixels or servos as well as Grove / STEMMA / Qwiic I2C ports. This will allow us to build a wide range of audio/video AI projects while also easily inserting sensors and robotic devices!
The bottom is equipped with a controllable mini fan to cool the Raspberry Pi during intensive AI computations. Most importantly, there is a switch to completely disable the audio codec, so when it is off, your voice cannot be heard, protecting your privacy.
Voice Bonnet
Adafruit Voice Bonnet for Raspberry Pi
Next, we have the Adafruit Voice Bonnet for Raspberry Pi: two speakers and two microphones. Your Raspberry Pi acts like an electronic brain, and the Adafruit Voice Bonnet is like the mouth and ears of a human! With two microphones and two 1Watt speakers using high-quality I2S codecs, it can be used with any Raspberry Pi that has a 2×20 GPIO header, from Raspberry Pi Zero to Raspberry Pi 4.
The onboard WM8960 codec uses I2S digital audio for high-quality recording and playback, so it sounds much better than the headphone jack on Raspberry Pi (or the lack of a headphone jack on Raspberry Pi Zero). We placed ferrite beads and filtering capacitors on every input and output to achieve the best performance at the lowest cost.
We specifically designed this voice engine to be used when making machine learning projects (like DIY voice assistants). Please see the guide on creating a DIY Google Assistant.
https://blog.adafruit.com/2020/09/24/adafruit-voice-bonnet-makes-for-a-tricked-out-google-assistant-raspberrypi-adafruit-googleassistant-raspberry_pi-ml-tinyml-machinelearning/
You can perform various voice-activated or voice-recognition projects. With two microphones, localization can be achieved.