Speech emotion recognition based on Graph-LSTM neural network

Li, Yan; Wang, Yapeng; Yang, Xu; Im, Sio-Kei

doi:10.1186/s13636-023-00303-9

Cited by 5 publications

(1 citation statement)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Moreover, Amir and Tanaya [20] proposed a depth graph learning method that utilized GCN to learn speech features and perform emotion classification by modeling speech signals as graph data. Based on GCN, Li et al [21] introduced an LSTM aggregator and weighted pooling approach to achieve better performance than existing graph learning baselines. Although GCN captures the sequence information lost by CNN in the convolution process, the performance of GCN is not as good as that of CNN in local feature extraction.…”

Section: Introductionmentioning

confidence: 99%

Energy Efficient Graph-Based Hybrid Learning for Speech Emotion Recognition on Humanoid Robot

Wu,

Xu,

Seng

et al. 2024

Electronics

View full text Add to dashboard Cite

This paper presents a novel deep graph-based learning technique for speech emotion recognition which has been specifically tailored for energy efficient deployment within humanoid robots. Our methodology represents a fusion of scalable graph representations, rooted in the foundational principles of graph signal processing theories. By delving into the utilization of cycle or line graphs as fundamental constituents shaping a robust Graph Convolution Network (GCN)-based architecture, we propose an approach which allows the capture of relationships between speech signals to decode intricate emotional patterns and responses. Our methodology is validated and benchmarked against established databases such as IEMOCAP and MSP-IMPROV. Our model outperforms standard GCNs and prevalent deep graph architectures, demonstrating performance levels that align with state-of-the-art methodologies. Notably, our model achieves this feat while significantly reducing the number of learnable parameters, thereby increasing computational efficiency and bolstering its suitability for resource-constrained environments. This proposed energy-efficient graph-based hybrid learning methodology is applied towards multimodal emotion recognition within humanoid robots. Its capacity to deliver competitive performance while streamlining computational complexity and energy efficiency represents a novel approach in evolving emotion recognition systems, catering to diverse real-world applications where precision in emotion recognition within humanoid robots stands as a pivotal requisite.

show abstract

Section: Introductionmentioning

confidence: 99%

Energy Efficient Graph-Based Hybrid Learning for Speech Emotion Recognition on Humanoid Robot

Wu,

Xu,

Seng

et al. 2024

Electronics

View full text Add to dashboard Cite

show abstract

PCQ: Emotion Recognition in Speech via Progressive Channel Querying

Wang,

et al. 2024

Lecture Notes in Computer Science

View full text Add to dashboard Cite

A Study on Synergistic Enhancement of Product Directive Optimization Strategies and Export Trade Performance in the Digital Era

Wang,

Dai

2024

Applied Mathematics and Nonlinear Sciences

View full text Add to dashboard Cite

The topic of innovation appears again and again in the government work report. The state and enterprises also pay more and more attention to product optimization and the output of innovative products, and more innovative and technological content has emerged from the products. At the same time, China’s foreign trade is transitioning from the traditional big country of foreign trade to the modern foreign trade power. In this paper, we combine natural language understanding and interactive image command mapping networks to improve the command optimization of digital products. The improved Graph-LSTM model is utilized to process the natural language instruction vectors, and the ICF network is utilized to take the environment image and user instructions as inputs to construct a fusion network in which the environment image and the natural language instructions can complement each other to complete the instruction mapping and to generate the location heat map for the execution of the instructions. Using principal component analysis, panel regression modeling, and other methods, we extracted and screened indicators of the influencing factors of export trade performance evaluation and constructed the performance evaluation index system. The synergistic enhancement of product instruction optimization and export trade performance is analyzed through empirical analysis and enhancement effect evaluation. The regression results of the selection equation are analyzed, and in the second column of the model, the coefficients of marketing cost and financial cost are 35.254 and -2.568, respectively, and the P-value is greater than 0.1, which can be analyzed in the next stage. In terms of the level of economic development, the difference between the comprehensive performance of export trade between Guangdong and Heilongjiang is 55.11, which is a large gap, and the optimization of product instructions has a greater impact on the performance of export trade in the eastern region.

show abstract

Speech emotion recognition based on Graph-LSTM neural network

Cited by 5 publications

References 46 publications

Energy Efficient Graph-Based Hybrid Learning for Speech Emotion Recognition on Humanoid Robot

Energy Efficient Graph-Based Hybrid Learning for Speech Emotion Recognition on Humanoid Robot

PCQ: Emotion Recognition in Speech via Progressive Channel Querying

A Study on Synergistic Enhancement of Product Directive Optimization Strategies and Export Trade Performance in the Digital Era

Contact Info

Product

Resources

About