Commit c03ac513 authored by Anika Treffehn's avatar Anika Treffehn
Browse files

Merge branch 'background_noise_modification' into 'main' 

Background noise modification

See merge request !23
parents a25aa301 989878fb

README.md

+39 −10
Original line number Diff line number Diff line
@@ -108,6 +108,7 @@ conditions_to_generate: 
          bin: ~/git/ivas-codec/IVAS_dec

postprocessing:

    fmt: "BINAURAL"

	fs: 48000

```



</details>
@@ -130,6 +131,8 @@ postprocessing: 
### Deletion of temporary directories containing 

### intermediate processing files, bitstreams etc.; default = false

# delete_tmp: true

### Master seed for random processes like bitstream error pattern generation; default = 0

# master_seed: 5



### Any relative paths will be interpreted relative to the working directory the script is called from!

### Usage of absolute paths is recommended.
@@ -160,13 +163,6 @@ output_path: "./tmp_output" 
### searches for the specified substring in found filenames; default = null

# input_select:

#  - "48kHz"



### Horizontally concatenate input items into one long file; default = false

# concatenate_input: true

### Specify preamble duration in ms; default = 0

# preamble: 40

### Flag wheter to use noise (amplitude +-4) for the preamble or silence; default = false (silence)

# pad_noise_preamble: true

```



</details>
@@ -220,6 +216,37 @@ input: 


</details>



### Optional pre-processing on whole signal(s)



<details>

<summary>Click to expand</summary>



```yaml

# preprocessing_2:

    ### Options for processing of the concatenated item (concatenate_input: true) or

    ### the individual items (concatenate_input: false) after previous pre-processing step

    ### Horizontally concatenate input items into one long file; default = false

    # concatenate_input: true

    ### Specify the concatenation order in a list of strings. If not specified, the concatenation order would be

    ### as per the filesystem on the users' device

    ### Should only be used if concatenate_input = true

    # concatenation_order: []

    ### Specify preamble duration in ms; default = 0

    # preamble: 10000

    ### Flag wheter to use noise (amplitude +-4) for the preamble or silence; default = false (silence)

    # preamble_noise: true

    ### Additive background noise

    # background_noise:

        ### REQUIRED: SNR for background noise in dB

        # snr: 10

        ### REQUIRED: Path to background noise, must have same format and sampling rate as input signal(s)

        # background_noise_path: ".../noise.wav"

        ### Seed for delay offest; default = 0

        # seed_delay: 10

```



</details>



### Bitstream processing



<details>
@@ -248,6 +275,8 @@ input: 
    # error_pattern: "path/pattern.192"

    ### Error rate in percent

    # error_rate: 5

	### Additional seed to specify number of preruns; default = 0

    # prerun_seed: 2

```

</details>
@@ -341,7 +370,7 @@ conditions_to_generate: 
          ### Path to decoder binary; default search for IVAS_dec in bin folder (primary) and PATH (secondary)

          bin: ~/git/ivas-codec/IVAS_dec

          ### Decoder output format; default = postprocessing fmt

          fmt: "CICP19"

          fmt: "7_1_4"

          ### Decoder output sampling rate; default = null (same as input)

          # fs: 48000

          ### Additional commandline options; default = null
@@ -382,8 +411,8 @@ conditions_to_generate: 
postprocessing:

    ### REQUIRED: Target format for output

    fmt: "BINAURAL"

    ### Target sampling rate in Hz for resampling; default = null (no resampling)

    # fs: 16000

    ### REQUIRED: Target sampling rate in Hz for resampling; default = null (no resampling)

    fs: 48000

    ### Low-pass cut-off frequency in Hz; default = null (no filtering)

    # lp_cutoff: 24000

    ### Target loudness in LKFS; default = null (no loudness change applied)

examples/TEMPLATE.yml

+1 −1
Original line number Diff line number Diff line
@@ -106,7 +106,7 @@ input: 
    # background_noise:

        ### REQUIRED: SNR for background noise in dB

        # snr: 10

        ### REQUIRED: Path to background noise

        ### REQUIRED: Path to background noise, must have same format and sampling rate as input signal(s)

        # background_noise_path: ".../noise.wav"

        ### Seed for delay offest; default = 0

        # seed_delay: 10

ivas_processing_scripts/__init__.py

+7 −0
Original line number Diff line number Diff line
@@ -45,6 +45,7 @@ from ivas_processing_scripts.processing import chains, config 
from ivas_processing_scripts.processing.processing import (

    preprocess,

    preprocess_2,

    preprocess_background_noise,

    process_item,

    reorder_items_list,

    reverse_process_2,
@@ -127,12 +128,18 @@ def main(args): 


        # run preprocessing only once

        if hasattr(cfg, "preprocessing"):

            # save process info for background noise

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

            preprocess(cfg, logger)



        # preprocessing on whole signal(s)

        if hasattr(cfg, "preprocessing_2"):

            # 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"):

                preprocess_background_noise(cfg)

            # preprocess 2

            preprocess_2(cfg, logger)



        # run conditions

ivas_processing_scripts/processing/chains.py

+1 −0
Original line number Diff line number Diff line
@@ -139,6 +139,7 @@ def get_preprocessing_2(cfg: TestConfig) -> dict: 
            "seed_delay": background_cfg.get("seed_delay", 0),

            "master_seed": cfg.master_seed,

            "output_fmt": cfg.postprocessing["fmt"],

            "background_object": None,

        }

    else:

        background = None

ivas_processing_scripts/processing/preprocessing_2.py

+10 −8
Original line number Diff line number Diff line
@@ -32,7 +32,6 @@ 


import logging

from pathlib import Path

from warnings import warn



import numpy as np
@@ -104,6 +103,9 @@ class Preprocessing2(Processing): 
        range_delay = (1, 2400000)



        # load background noise

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

            noise_object = self.background_noise["background_object"]

        else:

            noise_object = audio.fromfile(

                self.in_fmt,

                self.background_noise["background_noise_path"],
@@ -115,7 +117,7 @@ class Preprocessing2(Processing): 
        if len(noise_object.audio) < len(audio_object.audio):

            raise ValueError("Background noise too short for audio signal")

        if len(noise_object.audio) - range_delay[1] < len(audio_object.audio):

            warn(

            raise ValueError(

                "Background noise may be to short for audio signal when considering the random delay"

            )