Shakthi_1.1v — GUI-Based Voice Authentication System

Shakthi_1.1v is a Python-powered voice authentication system that protects access to your system using your unique voice. It captures a voice input, compares it to stored samples using MFCC + Dynamic Time Warping, and grants access only when your voice matches.Made by Arvind-NITCG

This is not a general-purpose voice recognition tool - it is personalized. It's built to recognize you, and only you.

---

Features

• Records your voice via microphone
• Filters and cleans the audio
• Extracts MFCC features and compares using DTW (Dynamic Time Warping)
• Allows access only when the voice matches your stored samples
• Falls back to PIN (optional) on repeated failure
• Fully local and offline - no cloud or external API
• Can be converted into an .exe for deployment

---

Project Structure

Shakthi_1.1v/ —GUI_Voice_auth.py # Main GUI interface ├── record_pass.py # Used to record your voice samples ├── verify_pass.py # Auth engine using DTW + MFCC ├── Voice_test.wav # Voice input placeholder (used at runtime) ├── samples_folder/ # Stores your personal reference samples │ └── .gitkeep # Keeps the folder in version control ├── Shakthi_1.1.spec # PyInstaller spec file to build .exe ├── requirements.txt # Python package dependencies ├── .gitignore # To exclude build files and logs

---

⚙️ Setup Instructions

1️⃣ Clone the Repository

git clone https://github.com/Arvind-NITCG/Shakthi_1.1v.git
cd Shakthi_1.1v

---

Create Conda Environment (Recommended)

->We recommend using a Conda environment to manage dependencies, especially if you plan to build the .exe or run the full voice pipeline (librosa, scipy, noisereduce) However, using pip in a virtualenv or system Python also works.

-> Miniconda Installation (Recommended for Developers) We recommend using Miniconda to create an isolated Python environment for this project. It helps prevent package conflicts and makes it easier to manage dependencies like scipy and librosa.

Download Miniconda here:
👉 https://docs.conda.io/en/latest/miniconda.html create a virtual environment in conda like: conda create -n shakthi_env python=3.11 conda activate shakthi_env

Install Dependencies

    pip install -r requirements.txt

Having issues installing packages like scipy or sounddevice on Windows? You can download pre-built .whl files from Gohlke's Unofficial Binaries: https://www.lfd.uci.edu/~gohlke/pythonlibs/

Install them using: pip install path/to/downloaded_file.whl

Setting Up Your Voice

Before running the GUI, you must register your voice samples.

-> Step 1: Run record_pass.py This will let you record your own voice samples. python record_pass.py

The .wav files will be saved inside samples_folder/. These files act as your voice "password". Try to make atleast 6 samples of your voice for DTW comparison

⚠️ Before using the app, you must tune the DTW cost threshold for your own voice and make your own SAFE EXIT PIN

How to Find Your Threshold

1. Run record_pass.py and create your voice samples in samples_folder/.
2. Temporarily run `verify_pass.py manually several times using your own voice: python verify_pass.py 3.Observe the Output DTW Total Distance: 64000 Average cost per frame: 31.25
3. Set your COST_THRESHOLD slightly above your best matching average. Example: If your average cost is 31.25, try COST_THRESHOLD = 35.
4. Save the new value inside: COST_THRESHOLD = 35 # <- set in verify_pass.py
5. Open GUI_Voice_auth.py and in the line 45 set your Safe Exit Pin
6. You're now ready to run the GUI.

Once COST_THRESHOLD and Safe-exit pin is made ready , Run th GUI

python GUI_Voice_auth.py

-> Speak when prompted.

-> Your voice is recorded and saved to Voice_test.wav.

-> It is then compared against your samples in samples_folder/.

-> Acess is granted only if DTW cost is within the threshold.

Optional: Build Your Own .exe (for Personal Use or Deployment)

This project includes a Shakthi_1.1.spec file that you can use to build a standalone .exe version of the app — ideal if you want to auto-launch it on boot or deploy it on another system.

✅ Why Use It?

• Certain dependencies had to be hard coded in the .spec without this it may not be working
• Works well with Task Scheduler (boot protection)

🧱 Prerequisites

• Use a Conda environment (recommended for stable builds)
• Ensure your voice samples are ready in samples_folder/
• Set your own COST_THRESHOLD before building

📦 Build Instructions

🧩 Updating DLL Paths in .spec File (Important for .exe Build)

Your system may store important audio-related DLLs in a different location than ours. You must update these paths in the Shakthi_1.1.spec file before building the .exe.

🔍 How to Find the Correct DLL Paths

If you're using Conda (recommended):

1. Activate your environment: conda activate shakthi_env

2. Find the folder containing libportaudio-64bit.dll and other DLLs as mentioned in the .spec In conda environment type : where libportaudio-64bit.dll similarly find the locations of all important DLL and add them to binaries section of .spec by removing the existing DLL which is ours.

3. Also check the datas=[] section if you're including data like: datas=[ ('samples_folder', 'samples_folder'), ('Voice_test.wav', '.'), ]

⚠️ Warning: Do not place DLL files in the datas=[] section of the .spec file. All .dll files must go in binaries=[] so PyInstaller can correctly link them during runtime.

Building .exe

1. Activate your Conda environment: conda activate shakthi_env
2. Run pyinstaller pyinstaller Shakthi_1.1.spec 3.Your .exe will be created in the dist/ folder: dist/Shakthi_1.1/Shakthi_1.1.exe 4.You can now run the .exe or add it to Task Scheduler (see setup_task_scheduler.md).

📎 Note: We do not distribute the .exe file. Each user should build their own using their voice samples.

💡 Notes -> Voice_test.wav must exist - the GUI writes into it. Do not delete this file.

-> The system only recognizes your voice. If someone else uses it, authentication will fail.

-> The samples_folder/ is user-specific - others must create their own samples.

---

🤝 Contributions This version is built for personal use. Future versions (Shakthi_2.0 and beyond) may support multi-user or AI-based recognition. Feedback and suggestions are welcome.

📄 License This project is open-source under the MIT License.