Feature extraction

Cross-platform presence: All of our technological solutions have been customized for specific audio formats, signal degradations and computation capabilities for different platforms - IVR, Android, iOS, Web, and desktop (Mac OS).

Feature extraction

Our high-level voice-processing technologies (singing rating, voice suppression, mixing with effects) rely on very solid, robust, low-level acoustic-feature extraction algorithms. These have been optimized to work in real-time i.e. as the user's voice is being recorded.

Features

  • High-resolution voice-pitch extraction (every 10 ms) with octave-error correction
  • Voice-activity and silence detection
  • Vocal onset extraction at syllabic resolution
  • MFCC extraction for speech/lyric processing
  • Clipping detection