Real-Time Voice Pitch Detection on STM32F401RE

A real-time Digital Signal Processing (DSP) system for detecting the fundamental frequency (pitch) of the human voice using the Autocorrelation Method on the STM32F401RE microcontroller.

This project was developed for the Microprocessors and Assembly Language Course
(Academic Year 2025–2026).

---

📌 Project Overview

This system captures analog voice signals through the ADC peripheral of the STM32F401RE and processes them in real time to estimate the speaker’s pitch within the human vocal range (85 Hz – 350 Hz).

The detected frequency is mapped to the nearest musical note and displayed on a 16x2 Character LCD.

The implementation focuses on:

• Real-time signal acquisition
• Time-domain DSP processing
• Low-level embedded programming
• Efficient memory management
• Register-level peripheral control without HAL abstraction

---

Features

🔹 Real-Time Pitch Detection

Detects the fundamental frequency of human speech using the Autocorrelation Function (ACF).

🔹 Low-Level Peripheral Programming

LCD communication is implemented entirely using custom drivers in both:

• 4-bit mode
• 8-bit mode

No HAL LCD libraries or external drivers are used.

🔹 Circular Buffer (Ring Buffer)

Efficient FIFO-based sample storage for continuous ADC acquisition.

🔹 Signal Normalization

Input amplitudes are normalized before autocorrelation to improve resolution and stability.

🔹 Musical Note Mapping

Converts detected frequencies into corresponding musical notes such as:

• A4
• C3
• E2
• etc.

🔹 Proteus Simulation Support

Includes a complete Proteus project for simulation and testing.

---

Mathematical Background

The system uses the Autocorrelation Function to detect periodicity in quasi-periodic voice signals.

The autocorrelation of the discrete signal $x(n)$ is defined as:

$$ R_x(l) = \sum_{n \in Z} x(n)x(n-l) $$

The pitch period is determined by finding the first significant non-zero peak in the autocorrelation output.

The pitch frequency is then calculated using:

$$ f_{pitch} = \frac{f_s}{l_{peak}} $$

Where:

• $f_s$ → Sampling frequency
• $l_{peak}$ → Lag corresponding to the first autocorrelation peak

---

Hardware Specifications

Component	Description
Microcontroller	STM32F401RE (ARM Cortex-M4)
Display	16x2 LCD (HD44780 Compatible)
Input Source	Microphone / Function Generator
Development Tools	STM32CubeIDE / Keil MDK
Simulation	Proteus Design Suite

---

⚙️ Software Design

1️⃣ Sampling Stage

The ADC continuously samples the analog input signal and stores the data inside a circular FIFO buffer.

2️⃣ DSP Processing

The system computes the autocorrelation function over a fixed analysis window.

3️⃣ Peak Detection

Noise and insignificant peaks are ignored using threshold-based logic to identify the first valid autocorrelation peak.

4️⃣ Frequency Estimation

The lag value is converted into the estimated pitch frequency.

5️⃣ Musical Note Detection

The frequency is mapped to the nearest musical note.

6️⃣ LCD Display

Detected pitch and note information are displayed on the LCD in real time.

---

📂 Repository Structure

├── main core/
├── 4-bit LCD driver
├── Core DSP algorithms
│   ├── System initialization
│   └── Main application logic
│
├── 8bit mode LCD manipulation/
│   ├── 8-bit LCD driver
│   └── Register-level LCD routines
│
├── proteus project/
│   ├── Proteus schematic
│   └── Simulation files (.pdsprj)
│
├── audio files/
│   ├── Sample WAV files
│   └── Test audio signals
│
├── Final technical report
    └── Design specifications

Additional Documentation

The complete project description, implementation details, mathematical analysis, hardware design, and final report are available in the PDF documents included in the repository.