SUST Bangla Emotional Speech Corpus (SUBESCO): An audio-only emotional speech corpus for Bangla

Abstract

SUBESCO is an audio-only emotional speech corpus for Bangla language. The total duration of the corpus is in excess of 7 hours containing 7000 utterances, and it is the largest emotional speech corpus available for this language. Twenty native speakers participated in the gender-balanced set, each recording of 10 sentences simulating seven targeted emotions. Fifty university students participated in the evaluation of this corpus. Each audio clip of this corpus, except those of Disgust emotion, was validated four times by male and female raters. Raw hit rates and unbiased rates were calculated producing scores above chance level of responses. Overall recognition rate was reported to be above 70% for human perception tests. Kappa statistics and intra-class correlation coefficient scores indicated high-level of inter-rater reliability and consistency of this corpus evaluation. SUBESCO is an Open Access database, licensed under Creative Common Attribution 4.0 International, and can be downloaded free of charge from the web link: https://doi.org/10.5281/zenodo.4526477.

BanglaSER: A speech emotion recognition dataset for the Bangla language

“…To a degree, the neutral emotional class is mistaken for the sad class. Note that, similar confusion has been seen in prior studies too [ 2 , 11 ]. Recognition rates are highest for angry followed by neutral, happy, sad, and surprise emotions.…”

“…There are several SER datasets available in literature covering English, Greek, Korean, and German languages, such as IEMOCAP [5] , RAVDESS [6] , MSP-IMPROV [2] , AESDD [7] , SAVEE [8] , CADKES [9] , and EMO-DB [10] . However, there is only one dataset for the SER task in the Bangla language, namely SUBESCO [11] . A simple comparison of these public SER datasets with BanglaSER is shown in Table 1 .…”

Chung-Ang Auditory Database of Korean Emotional Speech: A Validated Set of Vocal Expressions With Different Intensities

“…Most recently, another database, whose verification is based on perception tests and statistical analyses, came from the Bangla language. The SUST Bangla Emotional Speech Corpus (SUBESCO) [16] involves 20 actors portraying 7 emotions.…”

“…However, the relationship was not significant (p = . 16). The identification accuracy-naturalness relationship is captured in Fig.…”

“…In the study of speech emotion recognition, validated emotional speech databases constitute a crucial building block for developing and evaluating speech emotion recognizers [13]. To date, numerous emotional speech databases have been created in many languages, including Arabic [14], [15], Bangla [16], Mandarin Chinese [17], [18], Danish [19], English [20]- [22], German [23], [24], Italian [25], and Persian [26]. However, the majority of the databases come from highresource languages such as English, Mandarin Chinese, and German [27].…”