Building 45 Part 8: Wiring and Program Planning
I Wanted to Get More Done this Week
It was a hard school week!
I had three tests so there wasn’t much time to work on the project. I really only had Saturday free so I tired to make the most of it. I mostly have the wiring done and I have the program planned out. I also tweaked the design a little bit to add detail. I think it looks very nice but the physical limitations of the design make that irrelevant. Sadly this model will be nothing more then a fancy desk ornament
I found out too late that I2S is different from I2C. I bought a USB sound card for the audio output and I’ll wire that in soon. The module I got is just an amp and doesn’t transmit the audio over I2C. I will leave it in though as it gives me greater control over the output.
Hardware Overview and Thoughts
It might fly at max throttle but I’m not excited to test that
I had my doubts in the first post but I pushed ahead anyway because I knew this model is for learning. I really thought that this could work initially which tells you how little I knew about drones. I gave it my best shot but the circuit and motors were just too heavy. I’m sure that using lighter motors and optimizing the design could make it fly but I think lighter then air is the only way. I plan to combine all the sensors onto one board for the next model to make it as light as possible. If I want to build a robot that isn’t annoying to be around then it will need to be as quiet as possible. This definitely isn’t it:
Speaking of Programming
I have the programming all planned out and I think I can pull it off?
I found out that I have to travel to a wedding the week before the due date so my time will be limited. I found some code online that I might be able to adapt to get things working but things will be mostly up for me to figure out. I go back and forth on whether I should use a wake word like “Hey 45” to begin sending audio or if I should just use a simple voice activity detection algorithm.
Crunch Time!
My school load will only increase in the coming weeks so this might be my only window to get things done
This week was disappointing because I had tests for my three hardest classes. I really wanted to get to the programming this week but school won’t get off my back. I’ll try my hardest to get the AI side of things worked out over the next week as it will be the only thing I can show off. If I get more of this worked out now then the next model will be easier. I plan to constantly steam audio to the server in chunks because client side keyword detection is very difficult.
Next week the plan is to:
- Write a Python script for the client that sends recorded audio to the socket server in 30 second chunks
- Transcribe the .wav file in the input directory using faster whisper and check for the “45” keyword
- Use the HF transformers library to run sentiment analysis on the transcribed audio
- Send the transcribed text to the OpenAI compatible API
- Use the XTTS API to stream? the output text as audio back to the RPi through the socket server
- Wire up the USB audio
- Write a Python script for the client that sends recorded audio to the socket server in 30 second chunks