Project 9: Pitch Shifter / Time Stretcher (Breaking the Speed of Sound)

Build a basic phase-based pitch shifter or time stretcher with audible results.

Project Overview

Attribute Value
Difficulty Level 3: Advanced
Time Estimate 2-3 weeks
Main Language C
Alternative Languages Python, Rust, C++
Knowledge Area Phase and windowed processing
Tools Audio player, plotting tool
Main Book “Understanding Digital Signal Processing” by Richard G. Lyons

What you’ll build: A tool that changes pitch without changing tempo, or tempo without changing pitch, using short-time analysis.

Why it teaches DSP: You must understand phase, window overlap, and the limits of time-frequency manipulation.

Core challenges you’ll face:

  • Maintaining phase continuity between frames
  • Managing artifacts like echo or warbling
  • Tuning window size and overlap for clarity

Real World Outcome

You will be able to shift the pitch of a voice or stretch a sound without changing its pitch, with clear but imperfect results.

Example Output:

$ ./pitch_shift --semitones +3 --input voice.wav --output voice_shifted.wav
Frames: 320
Window: 1024
Output: voice_shifted.wav

Verification steps:

  • Compare original and processed audio
  • Listen for phase-related artifacts

The Core Question You’re Answering

“How can I change the pitch or duration of a signal while preserving its character?”

This project shows the practical cost of violating time-frequency limits.

Concepts You Must Understand First

Stop and research these before coding:

  1. Phase vocoder basics
    • Why does phase continuity matter between frames?
    • Book Reference: “Understanding Digital Signal Processing” by Richard G. Lyons, Ch. 11
  2. Window overlap and synthesis
    • How do overlap-add methods reconstruct the signal?
    • Book Reference: “The Scientist and Engineer’s Guide to DSP” by Steven W. Smith, Ch. 17
  3. Artifacts and tradeoffs
    • What causes smearing or metallic sounds?
    • Book Reference: “Understanding Digital Signal Processing” by Richard G. Lyons, Ch. 11

Questions to Guide Your Design

  1. Processing strategy
    • Will you adjust playback rate or phase progression?
    • How will you keep frames aligned during synthesis?
  2. Quality tuning
    • What window size gives the best balance for speech vs music?
    • How will you expose quality vs speed tradeoffs?

Thinking Exercise

Frame Alignment

Imagine you analyze a signal with 50 percent overlap and then reconstruct it. If you double the analysis hop size, what happens to time duration?

Questions while working:

  • Why does changing hop size affect time scaling?
  • How does pitch shifting differ from time stretching?

The Interview Questions They’ll Ask

Prepare to answer these:

  1. “What is a phase vocoder?”
  2. “Why does phase continuity matter?”
  3. “How does overlap-add reconstruction work?”
  4. “What artifacts are common in pitch shifting?”
  5. “How would you evaluate audio quality objectively?”

Hints in Layers

Hint 1: Starting Point Start with time stretching by changing the hop size between frames.

Hint 2: Next Level Track phase differences between frames and preserve them on synthesis.

Hint 3: Technical Details Use a consistent window and overlap to reduce discontinuities.

Hint 4: Tools/Debugging Test with simple tones and verify that pitch or duration changes as expected.

Books That Will Help

Topic Book Chapter
Phase vocoder “Understanding Digital Signal Processing” by Richard G. Lyons Ch. 11
Overlap-add “The Scientist and Engineer’s Guide to DSP” by Steven W. Smith Ch. 17
Audio artifacts “Understanding Digital Signal Processing” by Richard G. Lyons Ch. 11

Implementation Hints

  • Focus on clarity, not perfection, in the first version.
  • Use a test tone sweep to reveal artifacts.
  • Keep window and hop sizes configurable.

Learning Milestones

  1. First milestone: You can time-stretch without total distortion.
  2. Second milestone: You can shift pitch by a few semitones.
  3. Final milestone: You can explain the artifacts and how to reduce them.