DirAC reproduction quality depends on one dominant source per frequency band

Evidence

This implementation of DirAC has been proven to produce better perceptual quality in loudspeaker listening than other available techniques using the same microphone input (Vilkamo et al., 2009). An almost authentic reproduction is obtained if the acoustic scenario fits the DirAC signal model, where sound from one source dominates in one frequency band with potentially some diffuse sound arising from room reverberation. On the other hand, two broadband sources with temporally relatively smooth envelopes that arrive at the microphone from opposite directions cause some quality degradation (Laitinen, 2014). — Ville Pulkki, Matti Karjalainen, Communication Acoustics- An…, 1869

Categorization

• XREF: The DirAC model constraint (one dominant source per band) parallels beamforming/localization assumptions in speech separation and other time-frequency masking work.