12/17/2025 - ideologhy-code-hardware-pcb-cad-firmware

Project Journey – AI Voice Assistant

(15 Days | ~7–8 Hours Per Day)

Day 1 – Understanding the Problem & Feature Brainstorming

Spent most of the day understanding what an AI voice assistant should realistically do.
Wrote down ideas, removed over-complex features, and finalized a practical feature list.
Researched existing assistants to avoid copying and to find gaps.

Day 2 – Feature Freezing & Component Research

Converted the feature list into technical requirements.
Shortlisted components and checked availability, cost, and compatibility.
Changed microcontroller choice once due to memory limitations.

Day 3 – System Architecture & Block Diagram

Created a full system flow: voice input → processing → response → output.
Made multiple rough block diagrams before finalizing one.
Reviewed power requirements and signal flow carefully.

Day 4 – Firmware Logic Planning

Planned firmware structure in detail.
Wrote flowcharts and pseudocode instead of jumping into coding.
This saved time later during debugging.

Day 5 – Firmware Setup & Basic Tests

Set up the IDE and libraries.
Faced dependency errors and version conflicts.
By the end of the day, basic firmware was compiling correctly.

Day 6 – Audio Input Integration

Worked on microphone input.
Faced noise and low sensitivity issues.
Tried multiple configurations before getting stable audio capture.

Day 7 – Speech-to-Text Integration

Integrated speech-to-text service.
Initial results were inaccurate.
Spent time adjusting audio gain, sampling rate, and buffering.

Day 8 – AI Response & Logic Handling

Implemented AI processing and response logic.
Focused on making responses sound natural instead of robotic.
Handled edge cases like silence and unclear commands.

Day 9 – Text-to-Speech & Speaker Output

Integrated text-to-speech output.
Faced speaker distortion issues.
Solved it by improving power regulation and audio routing.

Day 10 – Breadboard Hardware Testing

Assembled complete hardware on a breadboard.
Experienced loose connections and random resets.
Rewired everything neatly and added decoupling capacitors.

Day 11 – PCB Schematic Design

Designed PCB schematic from the tested circuit.
Cross-checked pin mapping and power lines multiple times.
Corrected mistakes found during review.

Day 12 – PCB Layout & Revision

Worked on PCB layout carefully, especially audio lines.
Made spacing adjustments to reduce noise.
Revised the layout twice to improve reliability.

Day 13 – CAD Enclosure Design

Designed the enclosure using CAD software.
Initially underestimated speaker size.
Redesigned the enclosure to fit all components comfortably.

Day 14 – Full System Integration & Stress Testing

Mentally simulated PCB behavior and firmware flow together.
Tested long conversations, repeated commands, and power cycles.
Fixed small logic bugs and response delays.

Day 15 – Documentation, Demo & Final Review

Prepared project documentation and journey notes.
Recorded demo responses and test scenarios.
Reviewed the entire project like an evaluator would

12/18/2025 - BOM

HERE IS BOM OF THIS PROJECT

BOM

12/26/2025 11:10 AM - Recognition voice with esp32 and inmp441 microphone

so first i connent the inmp441 microphone accoding to this digram

then design the code Structure

In which I plan to use Google api key for speech recognition Then I design a code In which we are capturing the audio from Microphone Convert into Linear 16 format Then sending it To Google for STT

#here is the final result we get from google stt

but wait that the issue i face it take a alot of time for stt of appox 10-20 second
in next post i try to fix by upgrade the libary and use other service

12/26/2025 11:48 AM - Recognition voice with esp32 and inmp441 microphone #2

so in our pervious post we see the time issue so inn this post i fix it
so first i connent the inmp441 microphone accoding to this digram and this time i add a sd card to store the audio in .wav format

then design the code Structure is total different from previous
this time the audio is store in sd card when the button press
later it send it to deepgram with a small package of 4096 byte then it get process and we get a responce

once the code is ready i upload it in esp32

here is the result i get

now i get the result in 2-3 second
but wait here is a catch i it get disconnect from server after 10-20 min even i use the keep alive feature then also