Integrating sequence information in the audio-visual detection of word prominence in a human-machine interaction scenario

Schnall, Andrea; Heckmann, Martin

doi:10.21437/interspeech.2014-565

Cited by 6 publications

(2 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The results indicate that Kernel ESZSL and SYNC-OVO / SYNC-OVO (rand) perform well on the GEMEP corpus, while EXEM (1NNS) and SYNC-OVO / SYNC-OVO (rand) achieve better performance on the DEMoS corpus (Figure 3). As an additional investigation, we present the UA and macro F1-score (calculated through averaging recalls and precision-recall integration) [71], [73] results of the strategies of Kernel ESZSL, EXEM (1NNS), SYNC-OVO, and SYNC-OVO (rand). The F1-score results obey the tendency of the UAs (Table VII), indicating that the analysis on the UAs in this work is possible to be transferred onto the joint precisionaccuracy measurement.…”

Section: B Experimental Results: Strategic Comparisonmentioning

confidence: 99%

“…To select optimal parameters for each strategies, we utilise emotion-independent 5-fold and 3-fold Cross-Validation (CV) in grid-searching on the GEMEP and DEMoS corpora, respectively, which makes the validation set include samples from two emotions in each CV round. The measurement of Unweighted Accuracy (UA) is chosen in the CV rounds as the standard, calculated through averaging recalls [7], [16], [71].…”

Section: Experimental Setup For Learning Strategiesmentioning

confidence: 99%

See 1 more Smart Citation

Exploring Zero-Shot Emotion Recognition in Speech Using Semantic-Embedding Prototypes

Deng²,

Cummins

et al. 2022

IEEE Trans. Multimedia

View full text Add to dashboard Cite

Speech Emotion Recognition (SER) makes it possible for machines to perceive affective information. Our previous research differed from conventional SER endeavours in that it focused on recognising unseen emotions in speech autonomously through machine learning. Such a step would enable the automatic leaning of unknown emerging emotional states. This type of learning framework, however, still relied on manual annotations to obtain multiple samples of each emotion. In order to reduce this additional workload, herein, we propose a zero-shot SER framework employing a per-emotion semantic-embedding paradigm to describe emotions in zero-shot SER, instead of using the sample-wise descriptors. Aiming to optimise the relationship between emotions, prototypes, and speech samples, this framework includes two types of learning strategies: Sample-wise learning and emotion-wise learning. These strategies apply a novel learning process to speech samples and emotions, respectively, via specifically designed semantic-embedding prototypes. We verify the utility of these approaches by performing an extensive experimental evaluation on two corpora on three aspects, namely the influence of different types of learning strategies, emotional-pair comparison, and the selections of semantic-embedding prototypes and paralinguistic features. The experimental results indicate that it is applicable to use semantic-embedding prototypes for zero-shot emotion recognition in speech, despite the influence of choosing optimal strategies and prototypes.

show abstract

Section: B Experimental Results: Strategic Comparisonmentioning

confidence: 99%