Effective modelling of human expressive states from voice by adaptively tuning the neuro-fuzzy inference system

• DOI: 10.11591/ijai.v13.i1.pp185-194
• Publication Date: 2023-11-30T18:21:28Z
• Authors: Surjyo Narayana P...; Niharika Pattanaik; Hemanta Kumar Palo
• Keywords: Feature (linguistics)

<span lang="EN-US">This paper aims to develop efficient speech-expressive models using the adaptively tuning neuro-fuzzy inference system (ANFIS). The developed differentiate a high-arousal happiness state from low-arousal sadness benchmark Berlin (EMODB) database. proposed low-cost flexible algorithms are self-tunable and can address several vivid real-world issues such as home tutoring, banking, finance sectors, criminal investigations, psychological studies, call centers, cognitive biomedical sciences. work develops structures by formulating novel feature vectors comprising both time frequency information. features considered pitch (F0), standard deviation of (SDF0), autocorrelation coefficient (AC), log-energy (E), jitter, shimmer, harmonic noise ratio (HNR), spectral centroid (SC), roll-off (SR), flux (SF), zero-crossing rate (ZCR). alleviate curse dimensionality associated with frame-level extraction, extracted at utterance level. Several performance parameters have been computed validate individual models. Further, ANFIS tested for their efficacy in combinational platform. chosen complementary augmented indeed shown improved more available information revealed our results.</span>