Quantum-voice-classifier

KO | EN

gitlite — search

#python #react #llm #golang #android #machine-learning #deep-learning #linux #macos #nodejs #cli #vue #ios #openai

Quantum-voice-classifier

(★ 27)

Python and Streamlit application for voice classification using audio processing, MFCC features, and a quantum model with PennyLane.

Quantum-voice-classifier 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Quantum Voice Classifier

Python and Streamlit application for voice classification using audio processing, MFCC features, and a quantum machine learning model with PennyLane.

## Demo

## Overview

**Quantum Voice Classifier** is a local web application that records voice samples, processes audio signals, extracts MFCC features, and trains a quantum machine learning model to classify two spoken words.

This project combines signal processing, audio feature extraction, and a variational quantum circuit built with PennyLane.

## Features

- Record voice samples directly from the microphone.
- Train a custom voice classifier using two different spoken words.
- Extract MFCC audio features with Librosa.
- Normalize audio feature vectors for quantum encoding.
- Use a variational quantum circuit for classification.
- Save and load trained model weights.
- Test the trained model with new voice recordings.
- Display system status and signal-processing information.
- Visualize signal sampling and audio-processing flow.

## Technologies Used

- Python
- Streamlit
- Librosa
- PennyLane
- NumPy
- SoundDevice
- SciPy
- Matplotlib

## Project Structure

```text
Quantum-voice-classifier/
│
├── app.py
├── config.json
├── README.md
├── LICENSE
├── .gitignore
│
├── data/
│   ├── clase1.wav
│   ├── clase2.wav
│   ├── prueba.wav
│   └── temp audio files
│
├── models/
│   ├── pesos_entrenados.npy
│   └── pesos_encendidoapagado.npy
│
└── src/
    ├── final.py
    └── prueba1.py
```

## How It Works

1. The user records voice samples for two different words.
2. The audio signal is captured at 16,000 Hz.
3. The system checks if the recorded audio contains enough voice volume.
4. Silent sections are trimmed from the signal.
5. MFCC features are extracted from the audio.
6. The feature vector is normalized.
7. The normalized vector is encoded into a quantum circuit using Amplitude Embedding.
8. A variational quantum circuit processes the encoded data.
9. The model returns a quantum expectation value.
10. The final value is used to classify the voice sample into one of two classes.

## Installation

Clone the repository:

```bash
git clone https://github.com/Setzedich/Quantum-voice-classifier.git
cd Quantum-voice-classifier
```

Install the required dependencies:

```bash
pip install streamlit librosa pennylane sounddevice scipy matplotlib
```

## Usage

Run the application with:

```bash
python -m streamlit run app.py
```

After running the command, Streamlit will open the application in your browser.

## Main Application

The main file is:

```text
app.py
```

This file contains the Streamlit interface, the audio-processing functions, the quantum circuit, the training process, and the prediction system.

## Quantum Model

The project uses PennyLane to create a quantum classifier with a `default.qubit` device.

The model uses:

- `AmplitudeEmbedding` to encode the normalized audio features into quantum states.
- `BasicEntanglerLayers` as the variational ansatz.
- A Pauli-Z expectation value as the final output.

The output is interpreted as:

```text
Positive value  -> Class 1
Negative value  -> Class 2
```

## Signal Processing

The system uses MFCCs to represent the characteristics of the voice signal.

MFCC features are useful in speech-processing tasks because they summarize frequency patterns that are important for distinguishing human speech sounds.

The audio-processing pipeline includes:

```text
Audio Recording
      ↓
Signal Validation
      ↓
Silence Trimming
      ↓
MFCC Extraction
      ↓
Normalization
      ↓
Quantum Encoding
      ↓
Classification
```

## Model Files

Trained model weights are saved in:

```text
models/pesos_entrenados.npy
```

The class configuration is saved in:

```text
config.json
```

Audio samples are stored in:

```text
data/
```

## Example Workflow

1. Open the application.
2. Go to the training tab.
3. Enter two different words.
4. Record samples for each class.
5. Train the model.
6. Go to the testing tab.
7. Record a new voice sample.
8. The system predicts the class of the spoken word.

## Author

Developed by **Setzedich**.

## License

This project is licensed under the MIT License.

Quantum-voice-classifier

Quantum-voice-classifier 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Quantum-voice-classifier

Quantum-voice-classifier 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README