Crate loqa_voice_dsp 

Loqa Voice DSP - Shared DSP Library for Voice Analysis

This crate provides core DSP functionality for voice analysis including:

• Pitch detection (YIN algorithm)
• Formant extraction (LPC)
• FFT utilities
• Spectral analysis
• HNR (Harmonics-to-Noise Ratio) for breathiness analysis
• H1-H2 amplitude difference for vocal weight analysis

The crate is designed to be used by:

• Loqa backend (native Rust)
• VoiceFind mobile app (via FFI)

Re-exports

pub use analyzer::VoiceAnalyzer;

pub use config::AnalysisConfig;

pub use config::PitchAlgorithm;

pub use fft::compute_fft;

pub use fft::FFTResult;

pub use formants::extract_formants;

pub use formants::FormantResult;

pub use h1h2::calculate_h1h2;

pub use h1h2::H1H2Result;

pub use hnr::calculate_hnr;

pub use hnr::HNRResult;

pub use pitch::detect_pitch;

pub use pitch::PitchResult;

pub use pitch::PitchTrack;

pub use spectral::analyze_spectrum;

pub use spectral::SpectralFeatures;

Modules

analyzer

Stateful voice analyzer for frame-based and streaming analysis

autocorrelation

Normalized autocorrelation for pitch detection and voice analysis

config

Configuration for voice analysis

ffi

FFI (Foreign Function Interface) bindings for mobile platforms

fft

This module provides Fast Fourier Transform (FFT) computation for audio analysis. It wraps rustfft and realfft for efficient frequency domain transformations.

formants

Formant extraction using Linear Predictive Coding (LPC)

h1h2

H1-H2 amplitude difference calculation for vocal weight analysis

hnr

Harmonics-to-Noise Ratio (HNR) calculation for voice analysis

pitch

Pitch detection using YIN and autocorrelation algorithms

pyin

pYIN (Probabilistic YIN) pitch detection algorithm

spectral

This module provides spectral features for analyzing voice characteristics: