#include <stn_decomposer.hpp>
|
| void | prepare (const StnConfig &config) |
| | RT contract: prepare() allocates history, scratch, and mask storage and is not audio-thread safe.
|
| int | num_bins () const |
| const StnMasksT< SampleType > & | process (const SampleType *mag) |
| | Push one magnitude frame and compute its STN masks.
|
| const SampleType * | center_magnitude () const |
| | The center-frame magnitude the current masks apply to (delayed by (time_median-1)/2 frames vs the newest pushed frame).
|
| int | latency_frames () const |
| void | reset () |
◆ center_magnitude()
template<typename SampleType = float>
The center-frame magnitude the current masks apply to (delayed by (time_median-1)/2 frames vs the newest pushed frame).
◆ checked_retained_bytes()
template<typename SampleType = float>
◆ latency_frames()
template<typename SampleType = float>
◆ num_bins()
template<typename SampleType = float>
◆ prepare()
template<typename SampleType = float>
◆ process()
template<typename SampleType = float>
Push one magnitude frame and compute its STN masks.
mag holds num_bins magnitudes (linear). The returned masks reference internal storage valid until the next call. The time-median is centered on the frame that sits in the middle of the history, so there is an inherent (time_median-1)/2-frame latency before a frame's masks are final; transient material in the newest frame is still detected (the frequency-median is computed on the newest frame directly).
◆ reset()
template<typename SampleType = float>
The documentation for this class was generated from the following file: