Removal of spatial artifacts from audio

A non-transitory computer-readable medium to modify an audio stream comprising a plurality (n) of audio sources with a respective left (Lt) channel and a right (Rt) channel for each source, the audio sources being separated from an original audio stream, wherein the Lt channel and the Rt channel are in a time-frequency representation with instructions stored thereon that, when executed by one or more computers, cause the one or more computers to perform operations, the operations comprising, for each audio source (k):

• determining a left magnitude of the audio source (LSk) and a right magnitude of the audio source (RSk);
• determining an amplitude difference (Dk) between the LSk and the RSk;
• calculating a temporal derivative d(Dk) of the Dk;
• determining an average of LSk and RSk to obtain a mid-channel spectrogram (MCSk);
• normalizing the MCSk based on a sum of mid-channel spectrograms for each of the separated audio sources (MCS1+MCS2+ . . . +MCSn) to obtain a normalized value (Rk);
• dividing d(DK) by Rk to obtain a confidence map, wherein different regions of the confidence map are associated with respective likelihood values that indicate a respective likelihood that the region corresponds to a spatial artifact;
• computing a blending weight by scaling and clipping the confidence map;
• and
• combining the MCSk, the blending weight, and the Lt to obtain a left modified channel, and combining the MCSk, the blending weight, and the Rt to obtain a right modified channel.

A computing device to modify an audio stream comprising a plurality (n) of audio sources with a respective left (Lt) channel and a right (Rt) channel for each source, the audio sources being separated from an original audio stream, wherein the Lt channel and the Rt channel are in a time-frequency representation, the computing device comprising:

• a processor;
• and
• a memory coupled to the processor, with instructions stored thereon that, when executed by the processor, cause the processor to perform operations comprising, for each audio source (k):
• determining a left magnitude of the audio source (LSk) and a right magnitude of the audio source (RSk);
• determining an amplitude difference (Dk) between the LSk and the RSk;
• calculating a temporal derivative d(Dk) of the Dk;
• determining an average of LSk and RSk to obtain a mid-channel spectrogram (MCSk);
• normalizing the MCSk based on a sum of mid-channel spectrograms for each of the separated audio sources (MCS1+MCS2+ . . . +MCSn) to obtain a normalized value (Rk);
• dividing d(DK) by Rk to obtain a confidence map, wherein different regions of the confidence map are associated with respective likelihood values that indicate a respective likelihood that the region corresponds to a spatial artifact;
• computing a blending weight by scaling and clipping the confidence map;
• and
• combining the MCSk, the blending weight, and the Lt to obtain a left modified channel, and combining the MCSk, the blending weight, and the Rt to obtain a right modified channel.