First Upload

ASVP_ESD.py

@@ -0,0 +1,167 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Tue Apr 18 16:14:58 2023
+
+@author: lin.kinwahedward
+"""
+#------------------------------------------------------------------------------
+# Standard Libraries
+import datasets
+import numpy as np
+import os
+#------------------------------------------------------------------------------
+"""The Audio, Speech, and Vision Processing Lab - Emotional Sound Database (ASVP - ESD)"""
+
+_CITATION = """\
+@article{gsj2020asvpesd,
+  title={ASVP-ESD:A dataset and its benchmark for emotion recognition using both speech and non-speech utterances},
+  author={Dejoli Tientcheu Touko Landry and Qianhua He and Haikang Yan and Yanxiong Li},
+  journal={Global Scientific Journals},
+  volume={8},
+  issue={6},
+  pages={1793--1798},
+  year={2020}
+}
+"""
+
+_DESCRIPTION = """\
+ASVP-ESD
+"""
+
+_HOMEPAGE = "https://www.kaggle.com/datasets/dejolilandry/asvpesdspeech-nonspeech-emotional-utterances?resource=download-directory"
+
+_LICENSE = "CC BY 4.0"
+
+_DATA_URL = "https://drive.google.com/uc?export=download&id=1aKnr5kXgUjMB5MAhUTZmm3b8gjP8qA3O"
+
+
+id2labels = {
+    1: "boredom,sigh",
+    2: "neutral,calm",
+    3: "happy,laugh,gaggle",
+    4: "sad,cry",
+    5:"angry,grunt,frustration",
+    6: "fearful,scream,panic",
+    7: "disgust,dislike,contempt",
+    8: "surprised,gasp,amazed",
+    9: "excited",
+    10: "pleasure",
+    11: "pain,groan",
+    12: "disappointment,disapproval",
+    13: "breath"
+}
+
+#------------------------------------------------------------------------------
+# Define Dataset Configuration (e.g., subset of dataset, but it is not used here.)
+class ASVP_ESD_Config(datasets.BuilderConfig):
+    #--------------------------------------------------------------------------
+    def __init__(self, name, description, homepage, data_url):
+        
+        super(ASVP_ESD_Config, self).__init__(
+            name = self.name,
+            version = datasets.Version("1.0.0"),
+            description = self.description,
+        )
+        self.name = name
+        self.description = description
+        self.homepage = homepage
+        self.data_url = data_url
+#------------------------------------------------------------------------------
+# Define Dataset Class
+class ASVP_ESD(datasets.GeneratorBasedBuilder):
+    #--------------------------------------------------------------------------
+    BUILDER_CONFIGS = [ASVP_ESD_Config(
+        name = "ASVP_ESD",
+        description = _DESCRIPTION,
+        homepage = _HOMEPAGE,
+        data_url = _DATA_URL
+        )]
+    #--------------------------------------------------------------------------
+    '''
+        Define the "column header" (feature) of a datum.
+        3 Features:
+            1) path_to_file
+            2) audio samples
+            3) emotion label
+    '''
+    def _info(self):
+        
+        features = datasets.Features(
+                {
+                    "path": datasets.Value("string"),
+                    "audio": datasets.Audio(sampling_rate = 16000),
+                    "label": datasets.ClassLabel(
+                        names = [
+                            "boredom,sigh",
+                            "neutral,calm",
+                            "happy,laugh,gaggle",
+                            "sad,cry",
+                            "angry,grunt,frustration",
+                            "fearful,scream,panic",
+                            "disgust,dislike,contempt",
+                            "surprised,gasp,amazed",
+                            "excited",
+                            "pleasure",
+                            "pain,groan",
+                            "disappointment,disapproval",
+                            "breath"
+                        ])
+                }
+            )
+        
+        # return dataset info and data feature info
+        return datasets.DatasetInfo(
+            description = _DESCRIPTION,
+            features = features,
+            homepage = _HOMEPAGE,
+            citation = _CITATION,
+        )
+    #--------------------------------------------------------------------------
+    def _split_generators(self, dl_manager):
+        
+        dataset_path = dl_manager.download_and_extract(self.config.data_url)
+        
+        return [
+            datasets.SplitGenerator(
+                # set the whole dataset as "training set". No worry, can split later! 
+                name = datasets.Split.TRAIN,
+                # _generate_examples()'s parameters, thus name must match!
+                gen_kwargs = {
+                    "dataset_path": dataset_path
+                },
+            )
+        ]
+    #--------------------------------------------------------------------------
+    def _generate_examples(self, dataset_path):
+        '''
+            Get the audio file and set the corresponding labels
+        '''
+        key = 0
+        actors = np.arange(129)
+        for dir_name in actors:
+            #--------------------------------------------------------------------------
+            dir_path = dataset_path + "/ASVP_ESD/Speech/actor_" + str(dir_name)
+            for filename in os.listdir(dir_path):
+                if filename.endswith(".wav"):
+                    labels = filename[:-4].split("_")
+                    yield key, {
+                        "path": dir_path + "/" + filename,
+                        # huggingface dataset's will use soundfile to read the audio file
+                        "audio": dir_path + "/" + filename,
+                        "label": id2labels[int(labels[0])],
+                    }
+                    key += 1
+            #--------------------------------------------------------------------------
+            dir_path = dataset_path + "/ASVP_ESD/NonSpeech/actor_" + str(dir_name)
+            for filename in os.listdir(dir_path):
+                if filename.endswith(".wav"):
+                    labels = filename[:-4].split("_")
+                    yield key, {
+                        "path": dir_path + "/" + filename,
+                        # huggingface dataset's will use soundfile to read the audio file
+                        "audio": dir_path + "/" + filename,
+                        "label": id2labels[int(labels[0])],
+                    }
+                    key += 1
+#------------------------------------------------------------------------------

README.md

@@ -1,3 +1,41 @@
 ---
 license: cc-by-4.0
 ---
+
+# The Audio, Speech, and Vision Processing Lab - Emotional Sound Database (ASVP - ESD)
+
+## ABOUT
+
+The Audio, Speech, and Vision Processing Lab - Emotional Sound Database (ASVP - ESD) 
+was created .
+
+## CREATION OF THE DATABASE
+
+## CHOSEN EMOTIONS
+
+13 emotions were chosen:
+1. boredom,sigh
+2. neutral,calm
+3. happy,laugh,gaggle
+4. sad,cry
+5. angry,grunt,frustration
+6. fearful,scream,panic
+7. disgust,dislike,contempt
+8. surprised,gasp,amazed
+9. excited
+10. pleasure
+11. pain,groan
+12. disappointment,disapproval
+13. breath
+
+## ORGANISING THE DATABASE
+
+
+
+## References
+
+1. Dejoli Tientcheu Touko Landry, Qianhua He, Haikang Yan and Yanxiong Li. (2020). ASVP-ESD:A dataset and its benchmark for emotion recognition using both speech and non-speech utterances. Global Scientific Journals, 8(6), 1793-1798.
+
+
+
+