SER corpus import module

This module reads the wav files in DB folders to create clip file class objects All import functions read the DB folders in their original (downloaded as is from their sources) format.

create_DB_file_objects((string)db_name, (string)db_path) creates a list of file objected of all the wav files present in the db_path, as long as they are according to the formant of relevant 'db_name's original source and returns a list of Clip_file_Class objects.

# in example.py
import SER_DB_Parsing.SER_DB as SER_DB

list_of_clips = SER_DB.create_DB_file_objects("EmoDB", "path\\EMO-DB\\wav\\")

Each item in the list is created by the class: Clip_file_Class(db_id, filepath, speaker_id, scenario, sex, emotion_cat=None, intensity_cat=None, valance=None, arousal=None, dominance=None, naturalness=None, statement=None, repetition=None, n_raters=None, n_possible_emotions=None)

Get clip properties as:

print("First clip path:", list_of_clips[0].filepath)
print("First clip emotion category:", list_of_clips[0].emotion_cat)

Currently supports these DBs

db_name="EmoDB", db_path="C:\\DB\\EMO-DB\\wav\\"
# First file path: C:\DB\EMO-DB\wav\03a01Fa.wav

db_name="RAVDESS", db_path="C:\\DB\\RAVDESS\\"
# First file: C:\DB\RAVDESS\Speech\Actor_01\03-01-01-01-01-01-01.wav

db_name="IEMOCAP", db_path="C:\\DB\\IEMOCAP_noVideo\\"
# First file: C:\DB\IEMOCAP_noVideo\Session1\sentences\wav\Ses01F_script01_1\Ses01F_script01_1_F001.wav
# Evaluation file: C:\DB\IEMOCAP_noVideo\Session1\dialog\EmoEvaluation\Ses01F_impro01.txt

db_name="ShemoDB", db_path="C:\\DB\\shemo\\"
# First file: C:\DB\shemo\F\F01A01.wav

db_name="DEMoS", db_path="C:\\DB\\wav_DEMoS\\DEMOS\\"
# First file: C:\DB\wav_DEMoS\DEMOS\NP_f_01_col07b.wav

db_name="MSIMPROV", db_path="C:\\DB\\MSP-IMPROV\\"
# First file: C:\DB\MSP-IMPROV\session2\S01A\R\MSP-IMPROV-S01A-F02-R-FF01.wav
# Evaluation file: C:\DB\MSP-IMPROV\Evalution.txt

Change the db_path according to wherever you have stored the databases.

Standardized emotional category labels (single char) emotion_cat: {'N':'neutral', 'H':'happy', 'S':'sad', 'A':'anger', 'F':'fear', 'D':'disgust', 'U':'surprise', 'C':'calm', 'R':'frustuated', 'E':'excited', 'Y':'happy-excited', 'G':'guilty', 'X': 'unknown'}

Standardized scenario labels (int) : 0=unknown, 1=script, 2=improv, 3=radio/TV, 4=elicited, 5=natural, 6=script-in-improv

Standardized sexes (single char) : 'M'=males, 'F'=females

EmoDB http://emodb.bilderbar.info/docu/#docu

RAVDESS https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0196391

IEMOCAP https://sail.usc.edu/iemocap/iemocap_release.htm

ShEMO-DB https://github.com/pariajm/Persian-Emotional-Speech-Database-ShEMO

DEMoS https://zenodo.org/record/2544829

MSP-IMPROV https://ecs.utdallas.edu/research/researchlabs/msp-lab/MSP-Improv.html