📢 GSM 06.10 Full-Rate Speech Codec

🔥 Overview

This project implements a GSM 06.10 Full-Rate Speech Codec based on the Regular Pulse Excitation (RPE-LTP) algorithm. The codec is capable of encoding and decoding speech audio, transforming it into a compressed bitstream and reconstructing it back to speech.

🚀 Features

• ✅ Short-Term Linear Predictive Coding (LPC) for speech signal modeling
• ✅ Long-Term Prediction (LTP) for pitch tracking and residual analysis
• ✅ Regular Pulse Excitation (RPE) for efficient speech compression
• ✅ Bitstream Formation (260-bit frames) for transmission
• ✅ WAV file support for input and output
• ✅ Visualization of results via Matplotlib

📜 GSM 06.10 Standard

GSM 06.10 is a speech compression standard used in digital cellular systems. It processes audio in 160-sample frames (20ms at 8kHz sampling rate) and achieves full-rate speech encoding while maintaining quality suitable for telecommunication applications.

📁 Project Structure

📂 GSM-06.10-Codec
├── 📄 README.md        # This file
├── 📜 encoder.py       # Encoding process (LPC, LTP, RPE)
├── 📜 decoder.py       # Decoding process (inverse LPC, excitation reconstruction)
├── 📜 utils.py         # Helper functions for signal processing
├── 📜 hw_utils.py      # Utility functions for coefficient transformations
├── 📜 demo1.py         # Basic encoding/decoding demonstration
├── 📜 demo2.py         # Extended demonstration with enhanced features
├── 📜 demo3.py         # Full feature demonstration and performance evaluation
└── 🎵 ena_dio_tria.wav # Example input audio file

🛠 Installation & Setup

📌 Prerequisites

Make sure you have Python 3.x and the required dependencies installed:

pip install numpy scipy matplotlib bitstring

📌 Running the Encoder & Decoder

To process an audio file using the GSM 06.10 codec:

python demo3.py

This will:

1. Read the input .wav file
2. Encode it using GSM 06.10
3. Decode it back to a .wav file
4. Plot the original vs reconstructed waveform

🎯 How It Works

1️⃣ Short-Term Analysis

• Uses Linear Predictive Coding (LPC) to model speech
• Converts LPC coefficients to Log-Area Ratios (LAR)
• Encodes LAR using quantization

2️⃣ Long-Term Prediction

• Estimates pitch period (N) and gain factor (b)
• Predicts residuals based on prior speech frames

3️⃣ Regular Pulse Excitation (RPE)

• Selects sub-sequence with maximum energy
• Quantizes excitation signal
• Encodes into a 260-bit frame

4️⃣ Decoding Process

• Dequantizes & reconstructs excitation signal
• Applies inverse filtering to retrieve speech
• Synthesizes final speech waveform

📜 References

• GSM 06.10 Standard (ETSI)
• Python Signal Processing - SciPy