Function makeVoiceInMixLike

tests/fl/audio/detector/vocal.hpp:373–405

Generate a voice-in-mix-like signal matching real MP3 CQ features: centroid ~0.21, rolloff ~0.24, flatness ~0.46, density ~48 Guitar-like broadband base + strong harmonic overtones at voice fundamental. The harmonics create peaks in specific CQ bins, lowering flatness compared to the uniform guitar spectrum.

/// fundamental. The harmonics create peaks in specific CQ bins,
/// lowering flatness compared to the uniform guitar spectrum.
inline audio::Sample makeVoiceInMixLike(fl::u32 timestamp, float amplitude = 16000.0f,
                                       int count = 512, float sampleRate = 44100.0f) {
    fl::vector<fl::i16> data(count, 0);
    fl::fl_random rng(43);

    const float fMin = 150.0f;
    const float fMax = 5000.0f;

    // 1. Weak broadband base (much weaker than guitar → harmonics dominate)
    const int numComponents = 30;
    for (int c = 0; c < numComponents; ++c) {
        float t = static_cast<float>(rng.random16()) / 65535.0f;
        float freq = fMin * fl::powf(fMax / fMin, t);
        float envAmp = amplitude * 0.015f / fl::powf(freq / fMin, 1.1f);
        float phase0 = static_cast<float>(rng.random16()) / 65535.0f * 2.0f * FL_M_PI;
        for (int i = 0; i < count; ++i) {
            float phase = phase0 + 2.0f * FL_M_PI * freq * i / sampleRate;
            data[i] += static_cast<fl::i16>(envAmp * fl::sinf(phase));
        }
    }

    // 2. Very strong voice harmonics → dominate over broadband → lower flatness
    const float f0 = 180.0f;
    for (int h = 1; h * f0 < fMax; ++h) {
        float freq = f0 * h;
        float harmonicAmp = amplitude * 0.3f / fl::powf(static_cast<float>(h), 1.2f);
        for (int i = 0; i < count; ++i) {
            float phase = 2.0f * FL_M_PI * freq * i / sampleRate;
            data[i] += static_cast<fl::i16>(harmonicAmp * fl::sinf(phase));
        }
    }
    return audio::Sample(data, timestamp);
}

} // namespace test_vocal