John Lorenzo Bautista

Mr. John Lorenzo Bautista

University of Science and Technology, Daejeon, Republic of Korea
orcid

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
28 November 2022 in Electronics
Verified
Journal Article
Open access
PDF file
JBJohn Lorenzo Bautista
John Lorenzo Bautista
YLYun Kyung Lee
Yun Kyung Lee
HSHyun Soon Shin
Hyun Soon Shin
In this paper, an automatic speech emotion recognition (SER) task of classifying eight different emotions was experimented using parallel based networks trained using the Ryeson Audio-Visual Dataset of Speech and Song (RAVDESS) dataset. A combination of a CNN-based network and attention-based networks, running in parallel, was used to model both spatial features and temporal feature representations. Multiple Augmentation techniques using Additive White Gaussian Noise (AWGN), SpecAugment, Room Impulse Response (RIR), and Tanh Distortion techniques were used to augment the training data to further generalize the model representation. Raw audio data were transformed into Mel-Spectrograms as the model’s input. Using CNN’s proven capability in image classification and spatial feature representations, the spectrograms were treated as an image with the height and width represented by the spectrogram’s time and frequency scales. Temporal feature representations were represented by attention-based models Transformer, and BLSTM-Attention modules. Proposed architectures of the parallel CNN-based networks running along with Transformer and BLSTM-Attention modules were compared with standalone CNN architectures and attention-based networks, as well as with hybrid architectures with CNN layers wrapped in time-distributed wrappers stacked on attention-based networks. In these experiments, the highest accuracy of 89.33% for a Parallel CNN-Transformer network and 85.67% for a Parallel CNN-BLSTM-Attention Network were achieved on a 10% hold-out test set from the dataset. These networks showed promising results based on their accuracies, while keeping significantly less training parameters compared with non-parallel hybrid models.

10 January 2025 in Applied Sciences
Verified
Journal Article
Open access
PDF file
JBJohn Lorenzo Bautista
John Lorenzo Bautista
HSHyun Soon Shin
Hyun Soon Shin
This paper introduces a novel joint model architecture for Speech Emotion Recognition (SER) that integrates both discrete and dimensional emotional representations, allowing for the simultaneous training of classification and regression tasks to improve the comprehensiveness and interpretability of emotion recognition. By employing a joint loss function that combines categorical and regression losses, the model ensures balanced optimization across tasks, with experiments exploring various weighting schemes using a tunable parameter to adjust task importance. Two adaptive weight balancing schemes, Dynamic Weighting and Joint Weighting, further enhance performance by dynamically adjusting task weights based on optimization progress and ensuring balanced emotion representation during backpropagation. The architecture employs parallel feature extraction through independent encoders, designed to capture unique features from multiple modalities, including Mel-frequency Cepstral Coefficients (MFCC), Short-term Features (STF), Mel-spectrograms, and raw audio signals. Additionally, pre-trained models such as Wav2Vec 2.0 and HuBERT are integrated to leverage their robust latent features. The inclusion of self-attention and co-attention mechanisms allows the model to capture relationships between input modalities and interdependencies among features, further improving its interpretability and integration capabilities. Experiments conducted on the IEMOCAP dataset using a leave-one-subject-out approach demonstrate the model’s effectiveness, with results showing a 1–2% accuracy improvement over classification-only models. The optimal configuration, incorporating the joint architecture, dynamic weighting, and parallel processing of multimodal features, achieves a weighted accuracy of 72.66%, an unweighted accuracy of 73.22%, and a mean Concordance Correlation Coefficient (CCC) of 0.3717. These results validate the effectiveness of the proposed joint model architecture and adaptive balancing weight schemes in improving SER performance.