Commit 6a6e89e9 authored by Vladimir Malenovsky's avatar Vladimir Malenovsky
Browse files

fix incorrect overlap handling

parent a221268b

ivas_processing_scripts/generation/process_ism_items.py

+1 −1
Original line number Diff line number Diff line
@@ -193,7 +193,7 @@ def generate_ism_items( 
                N_delay = len(y.audio[:, 0])



                # add the shift

                N_delay += int(source_overlap * x.fs)

                N_delay += int(-source_overlap * x.fs)



                # ensure delay is a multiple of 20ms

                # N_delay = int(floor(source_shift * 50) / 50 * x.fs)

ivas_processing_scripts/generation/process_stereo_items.py

+13 −13
Original line number Diff line number Diff line
@@ -121,11 +121,11 @@ def generate_stereo_items( 
            )



            # convolve with stereo IR

            x_rev = reverb_stereo(x, IR)

            x = reverb_stereo(x, IR)



            # adjust the level of the stereo signal

            _, scale_factor = get_loudness(x_rev, cfg.loudness, "STEREO")

            x_rev.audio *= scale_factor

            _, scale_factor = get_loudness(x, cfg.loudness, "STEREO")

            x.audio *= scale_factor



            # shift the second (and all other) source files (positive shift creates overlap, negative shift creates a gap)

            if i > 0 and source_overlap != 0.0:
@@ -133,7 +133,7 @@ def generate_stereo_items( 
                N_delay = len(y.audio[:, 0])



                # add the shift

                N_delay += int(source_overlap * x.fs)

                N_delay += int(-source_overlap * x.fs)



                # ensure delay is a multiple of 20ms

                # N_delay = int(floor(source_shift * 50) / 50 * x.fs)
@@ -154,36 +154,36 @@ def generate_stereo_items( 
            # add source signal to the array of source signals

            y.fs = x.fs

            if y.audio is None:

                y.audio = x_rev.audio

                y.audio = x.audio

            else:

                # pad with zeros to have equal length of all source signals

                if x_rev.audio.shape[0] > y.audio.shape[0]:

                if x.audio.shape[0] > y.audio.shape[0]:

                    y.audio = np.vstack(

                        (

                            y.audio,

                            np.zeros(

                                (

                                    x_rev.audio.shape[0] - y.audio.shape[0],

                                    x.audio.shape[0] - y.audio.shape[0],

                                    y.audio.shape[1],

                                )

                            ),

                        )

                    )

                elif y.audio.shape[0] > x_rev.audio.shape[0]:

                    x_rev.audio = np.vstack(

                elif y.audio.shape[0] > x.audio.shape[0]:

                    x.audio = np.vstack(

                        (

                            x_rev.audio,

                            x.audio,

                            np.zeros(

                                (

                                    y.audio.shape[0] - x_rev.audio.shape[0],

                                    x_rev.audio.shape[1],

                                    y.audio.shape[0] - x.audio.shape[0],

                                    x.audio.shape[1],

                                )

                            ),

                        )

                    )



                # superimpose

                y.audio += x_rev.audio

                y.audio += x.audio



        # append pre-amble and post-amble to all sources

        if cfg.preamble != 0.0: