Skip to Main content Skip to Navigation

Distress Recognition from Speech Analysis: A Pairwise Association Rules-Based Approach

Abstract : Monitoring of the elderly has previously been done using visual, physiological devices, and physical nurse rounds. These have breached privacy and caused further health scares. The use of speech to read distress emotions is an alternative that, if utilized well, would provide effective monitoring while preserving the subject's privacy. Classical speech features have been used to feed machine learning algorithms to facilitate emotion detection. However, given the difference that individuals exhibit with their speech feature baselines with regard to the manifestation of distress on their speech given, it would make it difficult to normalize the features. The same case would apply for speech of an individual under different emotional circumstances. This paper proposes a novel approach where association rules drawn from speech features are used to derive the correlation between features and feed these correlations to machine learning techniques for distress detection. In achieving this, extraction of periodic segments is done, where each of these segments' features is paired with adjacent segments features, and their correlation percentages compared with other pairs within the same sound file, in order to establish a correlation between emotion features using association rules, creating defined rules that indicate specific emotions.
Document type :
Conference papers
Complete list of metadatas

Cited literature [28 references]  Display  Hide  Download
Contributor : Christophe Marsala <>
Submitted on : Thursday, December 12, 2019 - 3:19:56 PM
Last modification on : Wednesday, December 18, 2019 - 1:42:29 AM
Document(s) archivé(s) le : Friday, March 13, 2020 - 6:19:15 PM


Files produced by the author(s)


  • HAL Id : hal-02407488, version 1


Daniel Machanje, Joseph Orero, Christophe Marsala. Distress Recognition from Speech Analysis: A Pairwise Association Rules-Based Approach. IEEE Symposium Series on Computational Intelligence (SSCI) - Computational Intelligence for Engineering Solutions (CIES), Dec 2019, Xiamen, China. ⟨hal-02407488⟩



Record views


Files downloads