Commit a6d90601 authored by Archit Tamarapu's avatar Archit Tamarapu
Browse files

Merge branch... 

Merge branch '109-characterization-background-noise-mixing-for-osba-does-not-work-correctly' into 'main'

Resolve "[characterization] Background noise mixing for OSBA does not work correctly"

See merge request !215
parents 4a791e54 2ef8353c

ivas_processing_scripts/__init__.py

+1 −7
Original line number Diff line number Diff line
@@ -156,12 +156,6 @@ def main(args): 
            # save process info to revert it later

            cfg.pre2 = cfg.proc_chains[0]["processes"][0]

            # preprocess background noise

            if (

                hasattr(cfg, "preprocessing")

                and hasattr(cfg.pre2, "background_noise")

                and cfg.pre2.background_noise is not None

                and cfg.pre2.background_noise.get("background_noise_path")

            ):

            preprocess_background_noise(cfg)

            # preprocess 2

            preprocess_2(cfg, logger)

ivas_processing_scripts/audiotools/convert/__init__.py

+1 −1
Original line number Diff line number Diff line
@@ -427,7 +427,7 @@ def format_conversion( 
        if fmt.startswith("MASA"):

            output.metadata_file = input.metadata_file

        elif fmt.startswith("ISM"):  # also includes combined formats

            output.metadata_files = list(output.metadata_files)

            output.metadata_files = list(input.metadata_files)



    else:

        if isinstance(input, audio.BinauralAudio):

ivas_processing_scripts/audiotools/wrappers/filter.py

+48 −27
Original line number Diff line number Diff line
@@ -32,6 +32,7 @@ 


import re

from copy import deepcopy

from itertools import repeat

from pathlib import Path

from tempfile import TemporaryDirectory

from typing import Optional
@@ -43,11 +44,39 @@ from ivas_processing_scripts.audiotools.audio import Audio, ChannelBasedAudio 
from ivas_processing_scripts.audiotools.audioarray import delay_compensation, pad_delay

from ivas_processing_scripts.audiotools.audiofile import read, write

from ivas_processing_scripts.constants import DEFAULT_CONFIG_BINARIES

from ivas_processing_scripts.utils import find_binary, run

from ivas_processing_scripts.utils import apply_func_parallel, find_binary, run



FILTER_TYPES_REGEX = r"[\n][\s]{3}[A-Z0-9]\w+\s+"





def run_filter(

    cmd_base: list[str],

    audio: np.ndarray,

    fs: int,

    chan: int,

    skip_channel: list[int],

    block_size: Optional[int] = None,

):

    if chan in skip_channel:

        return audio



    with TemporaryDirectory() as tmp_dir:

        tmp_dir = Path(tmp_dir)

        tmp_in = tmp_dir.joinpath(f"tmpFilterOut_{chan}.pcm")

        tmp_out = tmp_dir.joinpath(f"tmpFilterIn_{chan}.pcm")



        cmd = cmd_base.copy()

        cmd.append(str(tmp_in))

        cmd.append(str(tmp_out))

        if block_size:

            cmd.append(str(block_size))



        write(tmp_in, audio, fs)

        run(cmd)

        out, _ = read(tmp_out, nchannels=1, fs=fs)

    return out





def filter_itu(

    input: Audio,

    flt_type: str,
@@ -155,32 +184,24 @@ def filter_itu( 
        # normal filtering -> size remains

        output = np.zeros_like(input.audio)



    with TemporaryDirectory() as tmp_dir:

        tmp_dir = Path(tmp_dir)



        # process channels separately

        for channel in range(input.num_channels):

            if skip_channel and channel in skip_channel:

                output[:, channel] = input.audio[:, channel]

                continue



            cmd_in_out = cmd.copy()



            tmp_in = tmp_dir.joinpath(f"tmp_filterIn{channel}.pcm")

            tmp_out = tmp_dir.joinpath(f"tmp_filterOut{channel}.pcm")



            cmd_in_out.append(str(tmp_in))

            cmd_in_out.append(str(tmp_out))



            if block_size:

                cmd_in_out.append(str(block_size))



            write(tmp_in, input.audio[:, channel], input.fs)



            run(cmd_in_out)

    # make sure this is an empty list

    skip_channel = skip_channel or []



    filtered = apply_func_parallel(

        run_filter,

        zip(

            repeat(cmd),

            [input.audio[:, ch] for ch in range(input.num_channels)],

            repeat(input.fs),

            range(input.num_channels),

            repeat(skip_channel),

            repeat(block_size),

        ),

        show_progress=False,

    )



            a, _ = read(tmp_out, nchannels=1, fs=input.fs)

            output[:, channel][:, None] = a

    for ch, filt in enumerate(filtered):

        output[:, [ch]] = filt.reshape(-1, 1)



    return output

ivas_processing_scripts/processing/preprocessing_2.py

+1 −1
Original line number Diff line number Diff line
@@ -135,7 +135,7 @@ class Preprocessing2(Processing): 
                raise ValueError(

                    f"Background noise path {self.background_noise.get('background_noise_path')} does not exist"

                )

            # load background noise

            # load preprocessed background noise

            if self.background_noise["background_object"] is not None:

                noise_object = self.background_noise["background_object"]

            else:

ivas_processing_scripts/processing/processing.py

+12 −3
Original line number Diff line number Diff line
@@ -569,7 +569,14 @@ def remove_preamble(x, out_fmt, fs, repeat_signal, preamble_len_ms, meta, logger 




def preprocess_background_noise(cfg):

    # TODO: add checks and errors for sampling rate and number channels compared to input signals

    if not (

        hasattr(cfg, "preprocessing")

        and hasattr(cfg.pre2, "background_noise")

        and cfg.pre2.background_noise is not None

        and cfg.pre2.background_noise.get("background_noise_path")

    ):

        return



    # create audio objects

    input_audio = audio.fromfile(

        cfg.input["fmt"],
@@ -583,7 +590,10 @@ def preprocess_background_noise(cfg): 
    else:

        out_fs = cfg.pre.out_fs



    # only consider format conversion, resampling and high-pass filtering

    if hasattr(input_audio, "object_pos"):

        output_audio.object_pos = input_audio.object_pos



    # only apply format conversion, resampling and filtering

    convert(

        input=input_audio,

        output=output_audio,
@@ -593,7 +603,6 @@ def preprocess_background_noise(cfg): 
    )



    # save result in cfg

    # cfg.preprocessing_2["background_noise"].update({"background_object": output_audio})

    cfg.proc_chains[0]["processes"][0].background_noise[

        "background_object"

    ] = output_audio