Commit be5bf7c5 authored by Archit Tamarapu's avatar Archit Tamarapu
Browse files

add support for rotation on input format (pre-rotation)

parent 17644dd4

ivas_processing_scripts/audiotools/__init__.py

+8 −8
Original line number Diff line number Diff line
@@ -96,13 +96,20 @@ def add_processing_args(group, input=True): 
        default=None,

    )

    group.add_argument(

        f"-{ps}t",

        f"-{ps}x",

        f"--{p}_trim",

        type=float,

        nargs=2,

        metavar=("PRE_TRIM", "POST_TRIM"),

        help="Pre-/post-trim the signal by this amount in milliseconds (negative values pad silence), (default = %(default)s)",

    )

    group.add_argument(

        f"-{ps}t",

        f"--{p}_trajectory",

        type=str,

        help="Head-tracking trajectory file for input pre-rotation or binaural output (default = %(default)s)",

        default=None,

    )

    group.add_argument(

        f"-{ps}pn",

        f"--{p}_pad_noise",
@@ -166,13 +173,6 @@ def get_args(): 
        help="Apply limiting to output (default = %(default)s)",

        action="store_true",

    )

    output_parser.add_argument(

        "-t",

        "--trajectory",

        type=str,

        help="Head-tracking trajectory file for binaural output (default = %(default)s)",

        default=None,

    )

    output_parser.add_argument(

        "-bd",

        "--bin_dataset",

ivas_processing_scripts/audiotools/convert/__init__.py

+4 −4
Original line number Diff line number Diff line
@@ -227,7 +227,7 @@ def process_audio( 
    esdru_alpha: Optional[float] = None,

    logger: Optional[logging.Logger] = None,

) -> None:

    """Perform (pre-/pos-) processing of audio"""

    """Perform (pre-/post-) processing of audio"""



    if fs is None:

        fs = x.fs
@@ -325,9 +325,9 @@ def format_conversion( 
    if logger:

        logger.debug(f"Format conversion: {input.name} -> {output.name}")



    if (fmt := input.name) == output.name or (

        input.name.startswith("BINAURAL") and output.name.startswith("BINAURAL")

    ):

    if (

        ((fmt := input.name) == output.name) and kwargs.get("in_trajectory") is None

    ) or (input.name.startswith("BINAURAL") and output.name.startswith("BINAURAL")):

        output.audio = input.audio

        if fmt.startswith("MASA"):

            output.metadata_file = input.metadata_file

ivas_processing_scripts/audiotools/convert/channelbased.py

+9 −3
Original line number Diff line number Diff line
@@ -57,9 +57,15 @@ from ivas_processing_scripts.audiotools.wrappers.filter import resample_itu 
def convert_channelbased(

    cba: audio.ChannelBasedAudio,

    out: audio.Audio,

    in_trajectory: Optional[Union[str, Path]] = None,

    **kwargs,

) -> audio.Audio:

    """Convert channel-based audio to the requested output format"""



    # pre-rotation if specified

    if in_trajectory is not None:

        cba.audio = rotate_cba(cba, in_trajectory)



    # CBA -> Binaural

    if isinstance(out, audio.BinauralAudio):

        render_cba_to_binaural(cba, out, **kwargs)
@@ -83,7 +89,7 @@ def convert_channelbased( 
def render_cba_to_binaural(

    cba: audio.ChannelBasedAudio,

    bin: audio.BinauralAudio,

    trajectory: Optional[Union[str, Path]] = None,

    out_trajectory: Optional[Union[str, Path]] = None,

    bin_dataset: Optional[str] = None,

    bin_lfe_gain: Optional[float] = None,

    **kwargs,
@@ -122,8 +128,8 @@ def render_cba_to_binaural( 
    cba.fs = 48000

    bin.fs = 48000



    if trajectory is not None:

        cba.audio = rotate_cba(cba, trajectory)

    if out_trajectory is not None:

        cba.audio = rotate_cba(cba, out_trajectory)



    IR, _, latency_smp = load_ir(cba.name, bin.name, bin_dataset)

ivas_processing_scripts/audiotools/convert/masa.py

+5 −4
Original line number Diff line number Diff line
@@ -71,7 +71,8 @@ def convert_masa( 
def render_masa_to_binaural(

    masa: audio.MetadataAssistedSpatialAudio,

    bin: audio.BinauralAudio,

    trajectory: Optional[Union[str, Path]] = None,

    in_trajectory: Optional[Union[str, Path]] = None,

    out_trajectory: Optional[Union[str, Path]] = None,

    bin_dataset: Optional[str] = None,

    **kwargs,

) -> None:
@@ -96,11 +97,11 @@ def render_masa_to_binaural( 


        render_masa_to_cba(masa, cba_tmp)



        channelbased.render_cba_to_binaural(cba_tmp, bin, trajectory)

        channelbased.render_cba_to_binaural(cba_tmp, bin, out_trajectory)

    else:

        if trajectory is not None:

        if in_trajectory is not None or out_trajectory is not None:

            warn(

                f"Head-rotation not supported by MasaRenderer! Trajectory {trajectory} will be ignored!"

                f"Head-rotation not supported by MasaRenderer! Trajectory {in_trajectory or out_trajectory} will be ignored!"

            )

        if bin_dataset is not None:

            warn(

ivas_processing_scripts/audiotools/convert/objectbased.py

+17 −5
Original line number Diff line number Diff line
@@ -60,10 +60,24 @@ from ivas_processing_scripts.utils import apply_func_parallel 
def convert_objectbased(

    oba: audio.ObjectBasedAudio,

    out: audio.Audio,

    in_trajectory: Optional[Union[str, Path]] = None,

    **kwargs,

) -> audio.Audio:

    """Convert an ISM signal to the requested output format"""



    # pre-rotation if specified - 20ms only!

    if in_trajectory is not None:

        # repeat each value four times since head rotation data is on sub-frame basis

        azi = np.repeat(oba.obj_pos[:, 0], 4)

        ele = np.repeat(oba.obj_pos[:, 1], 4)



        # apply head-rotation trajectory

        azi, ele = rotate_oba(azi, ele, in_trajectory)



        # update object metadata

        oba.obj_pos[:, 0] = azi[:, ::4]

        oba.obj_pos[:, 1] = ele[:, ::4]



    # OBA -> Binaural

    if isinstance(out, audio.BinauralAudio):

        render_oba_to_binaural(oba, out, **kwargs)
@@ -86,7 +100,7 @@ def convert_objectbased( 
def render_oba_to_binaural(

    oba: audio.ObjectBasedAudio,

    bin: audio.BinauralAudio,

    trajectory: Optional[Union[str, Path]] = None,

    out_trajectory: Optional[Union[str, Path]] = None,

    bin_dataset: Optional[str] = None,

    **kwargs,

) -> None:
@@ -105,15 +119,13 @@ def render_oba_to_binaural( 
        Name of binaural dataset, if None default dataset is used

    """



    # bin.audio = np.zeros([oba.audio.shape[0], bin.num_channels])



    if "ROOM" in bin.name:

        cba_tmp = audio.fromtype("7_1_4")

        cba_tmp.fs = oba.fs



        render_oba_to_cba(oba, cba_tmp)



        render_cba_to_binaural(cba_tmp, bin, trajectory)

        render_cba_to_binaural(cba_tmp, bin, out_trajectory)

    else:

        IR, SourcePosition, latency_smp = load_ir(oba.name, bin.name, bin_dataset)
@@ -130,7 +142,7 @@ def render_oba_to_binaural( 
                obj_idx,

                obj_pos,

                repeat(oba),

                repeat(trajectory),

                repeat(out_trajectory),

                repeat(IR),

                repeat(SourcePosition),

            ),