Equipe Analyse/Synthèse

Some descriptors will rely on textual attributes used by representative populations of subjects to qualify sound properties,

Others may rely on predefined sound classes such as sound sources.

Low-level descriptors and high-level descriptors will be mapped, based on various statistical analysis techniques:

• Multi-Feature StructuredVector Quantization (MFSVQ),
• High-order statistics (HOS),
• Independent Component Analysis (ICA),
• Agnostic matching of MFSVQ sequences,
• Hidden Markov Models (HMM) for dynamic sequences,

Page remise à jour le 14 Janvier 2001