Class STFT

include/engine/framework/audio/dsp.h:37–54

Speech-focused STFT surface used by voice models in this repo. It assumes real float input and returns one-sided complex or magnitude output.

// Speech-focused STFT surface used by voice models in this repo.
// It assumes real float input and returns one-sided complex or magnitude output.
class STFT {
public:
    AudioTensor compute_magnitude(
        const std::vector<float> & waveform,
        const std::vector<float> & window,
        int64_t batch,
        int64_t samples,
        const STFTConfig & config,
        size_t threads = 0) const;

    AudioTensor compute_complex(
        const std::vector<float> & waveform,
        const std::vector<float> & window,
        int64_t batch,
        int64_t samples,
        const STFTConfig & config,
        size_t threads = 0) const;
};

class ISTFT {
public: