Hybrid LSTM–Attention and CNN Model for Enhanced Speech Emotion Recognition DOI Creative Commons

Fazliddin Makhmudov,

Alpamis Kutlimuratov,

Young Im Cho

et al.

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(23), P. 11342 - 11342

Published: Dec. 5, 2024

Emotion recognition is crucial for enhancing human–machine interactions by establishing a foundation AI systems that integrate cognitive and emotional understanding, bridging the gap between machine functions human emotions. Even though deep learning algorithms are actively used in this field, study of sequence modeling accounts shifts emotions over time has not been thoroughly explored. In research, we present comprehensive speech emotion-recognition framework amalgamates ZCR, RMS, MFCC feature sets. Our approach employs both CNN LSTM networks, complemented an attention model, enhanced emotion prediction. Specifically, model addresses challenges long-term dependencies, enabling system to factor historical experiences alongside current ones. We also incorporate psychological “peak–end rule”, suggesting preceding states significantly influence emotion. The plays pivotal role restructuring input dimensions, facilitating nuanced processing. rigorously evaluated proposed utilizing two distinct datasets, namely TESS RAVDESS. empirical outcomes highlighted model’s superior performance, with accuracy rates reaching 99.8% 95.7% These results notable advancement, showcasing our system’s precision innovative contributions recognition.

Language: Английский

Synthesis of Optimal Correction Functions in the Class of Disjunctive Normal Forms DOI Creative Commons
Anvar Kabulov,

Abdussattar Baizhumanov,

Islambek Saymanov

et al.

Mathematics, Journal Year: 2024, Volume and Issue: 12(13), P. 2120 - 2120

Published: July 5, 2024

The paper proposes to consider individual heuristics as unreliably operating parts of the information processing system. In a separate case, several different are adopted solve same problem, and results obtained adjusted in certain way. this problems arise that close methodology synthesizing reliable circuits from unreliable elements or making collective expert decision. work solves problem constructing an optimal correction function based on control material; classes functions k-valued logic under monotonicity restrictions studied. A theorem completeness class monotonic for arbitrary k is proved, basis given proved constructed. corrector disjunctive normal forms solved.

Language: Английский

Citations

1
Hybrid LSTM–Attention and CNN Model for Enhanced Speech Emotion Recognition DOI Creative Commons

Fazliddin Makhmudov,

Alpamis Kutlimuratov,

Young Im Cho

et al.

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(23), P. 11342 - 11342

Published: Dec. 5, 2024

Emotion recognition is crucial for enhancing human–machine interactions by establishing a foundation AI systems that integrate cognitive and emotional understanding, bridging the gap between machine functions human emotions. Even though deep learning algorithms are actively used in this field, study of sequence modeling accounts shifts emotions over time has not been thoroughly explored. In research, we present comprehensive speech emotion-recognition framework amalgamates ZCR, RMS, MFCC feature sets. Our approach employs both CNN LSTM networks, complemented an attention model, enhanced emotion prediction. Specifically, model addresses challenges long-term dependencies, enabling system to factor historical experiences alongside current ones. We also incorporate psychological “peak–end rule”, suggesting preceding states significantly influence emotion. The plays pivotal role restructuring input dimensions, facilitating nuanced processing. rigorously evaluated proposed utilizing two distinct datasets, namely TESS RAVDESS. empirical outcomes highlighted model’s superior performance, with accuracy rates reaching 99.8% 95.7% These results notable advancement, showcasing our system’s precision innovative contributions recognition.

Language: Английский

Citations

1