Investigation of Spoken-Language Detection and Classification in Broadcasted Audio Content

Abstract

The current paper focuses on the investigation of spoken-language classification in audio broadcasting content. The approach reflects a real-word scenario, encountered in modern media/monitoring organizations, where semi-automated indexing/documentation is deployed, which could be facilitated by the proposed language detection preprocessing. Multilingual audio recordings of specific radio streams are formed into a small dataset, which is used for the adaptive classification experiments, without seeking-at this step-for a generic language recognition model. Specifically, hierarchical discrimination schemes are followed to separate voice signals before classifying the spoken languages. Supervised and unsupervised machine learning is utilized at various windowing configurations to test the validity of our hypothesis. Besides the analysis of the achieved recognition scores (partial and overall), late integration models are proposed for semi-automatically annotation of new audio recordings. Hence, data augmentation mechanisms are offered, aiming at gradually formulating a Generic Audio Language Classification Repository. This database constitutes a program-adaptive collection that, beside the self-indexing metadata mechanisms, could facilitate generic language classification models in the future, through state-of-art techniques like deep learning. This approach matches the investigatory inception of the project, which seeks for indicators that could be applied in a second step with a larger dataset and/or an already pre-trained model, with the purpose to deliver overall results.

Media Studies, Audiovisual Media Communications, and Generations: The Case of Budding Journalists in Radio Courses in Greece

“…How to use free and open-source software from and through the Internet to quickly and easily create sound/audio media spots (see Kotsakis et al 2020;Vryzas et al 2020;Kalliris et al 2019;Matsiola et al 2019;Nicolaou et al 2019;Tsipas, Nikolaos, LazarosVrysis, CharalamposDimoulas, and George Papanikolaou, 2015;Kotsakis et al 2012a;Kotsakis et al 2012b;Aguayo Gonzalez et al 2009;Salmon et al 2008;Dimoulas et al 2000).…”

Extending Radio Broadcasting Semantics through Adaptive Audio Segmentation Automations

“…Motivated by the results of a previous research on program-adaptive pattern analysis for Voice/Music/Phone [1] and Language Discrimination taxonomies [6], the presented methodology functions as an add-on module towards the formulation of a dynamic Generic Audio Classification Repository. Hence, following already adopted hierarchical classification strategies, new schemes were adapted based on clustering techniques, but also their combination with supervised training methods.…”

“…However, there are issues regarding the inhomogeneity of labeling meta-data, while in some cases, ground-truth training pairs are difficult to obtain (or are even completely unavailable). Hence, combinations of supervised, semi-supervised and unsupervised data mining algorithms are utilized to serve the specific necessities of various real-world multimedia semantics [1,[4][5][6][7][8][9].…”

HMM Based Language Identification from Speech Utterances of Popular Indic Languages Using Spectral and Prosodic Features

“…The first LID was designed and explored with acoustic features to build a dictionary of words using speech sound. Prosody features also play a vital role in recognizing language from speech signal [1,2]. Prosodic features like pitch, energy, stress are different in tonal language compared to non-tonal languages [3,4].…”

“…It is the task to recognize the language of utterance without knowing the details of speaker and language content. It identifies the languages of speech utterance based on only raw signal of speech utterance [1].…”