Speaker: Wiliam Fernando López Gavilánez. Abstract: Audio classifiers designed for deployment across diverse devices often face unforeseen conditions during inference, attributable to device-specific characteristics. These challenges stem from variations in microphone transfer functions or on-chip digital signal pre-processing, which result… Read More
Analyzing DiaPer EEND Speaker Diarization Models on the RTVE2022 Dataset
Speaker: Juan Ignacio Álvarez Trejos. Abstract: The task of speaker diarization has lately been successfully tackled with end-to-end neural diarization (EEND) models instead of modular cascaded ones. Among them, the very new EEND Perceiver-based attractors (DiaPer) comes with a light… Read More
Analysis of Speaker Label Matching for Diarization of Long Audios on RTVE2022 Dataset
Speaker: Laura Herrera Alarcón. Abstract: This study introduces an algorithm to match predicted speaker labels from short audio segments into a final prediction. This involves extracting an x-vector for each speaker in each segment and applying constrained Agglomerative Clustering to… Read More
Towards Efficient Conformer-based Sound Event Detection
Speaker: Sara Barahona Quirós. Abstract: The Conformer architecture has shown excellent performance in accurately classifying sound events but lacks temporal precision when predicting time boundaries. While increasing the length of the input sequences can mitigate this issue, it also increases… Read More
Automatic Speech Recognition in Dialectal Data (COSER)
Speaker: Clara Adsuar Ávila. Abstract: In this project, we address the importance of enhancing the accessibility and usefulness of Deep Learning technologies for non-standard speakers. From a linguistic perspective, rural Spanish areas are rich in dialectal variety. However, most technology… Read More
Evaluating Posterior Probabilities: Decision Theory, Proper Scoring Rules, and Calibration
Speaker: Daniel Ramos Castro. Abstract: Most machine learning classifiers are designed to output posterior probabilities for the classes given the input sample. These probabilities may be used to make the categorical decision on the class of the sample; provided as… Read More
One model to rule them all? Towards end-to-end joint speaker diarization and speech recognition
Speaker: Laura Herrera Alarcón. Abstract: This paper presents a novel framework for joint speaker diarization (SD) and automatic speech recognition (ASR), named SLIDAR (sliding-window diarization-augmented recognition). SLIDAR can process arbitrary length inputs and can handle any number of speakers, effectively… Read More
Emotion recognition in Spanish audio
Speaker: Manuel Otero González. Abstract: En esta charla se explicará la tarea de reconocimiento de emociones en audios en español, presentando los enfoques más avanzados del estado del arte, como Wav2Vec2 y W2V-Bert. Además, se introducirá el reto EmoSPeech, cuyo… Read More
State of the Art in Sound Event Detection and DCASE Evaluations
Speaker: Doroteo Torre Toledano. Abstract: In this talk I will review the most recent trajectory of the AUDIAS group in the field of Sound Event Detection (SED), highlighting our participations in DCASE evaluations (Task 4) from 2020 to 2023. Then,… Read More
Large Language Models: From Theory to Practice in Text Classification
Speaker: Miguel Ángel Martínez Pay. Abstract: This work presents a comprehensive overview of Large Language Models (LLMs), from their theoretical framework to practical applications in text classification. It compares the effectiveness of two key approaches: fine-tuning embeddings of smaller models… Read More