Extraction and Representation of Prosodic Features
for Automatic Speaker Recognition Technology
Journal:
GRENZE International Journal of Computer Theory and Engineering
Authors:
Nilu Singh, R. A. Khan
Volume:
2
Issue:
1
Grenze ID:
01.GIJCTE.2.1.11
Pages:
1-7
Abstract
To recognize emotion from a speech signal, the feature extraction technique used
is known as Prosodic features extraction technique. It is most common methodology to
emotion recognition depend on utterance level. Current studies shows that segmental
spectral features of a speech signal rely on utterance level measurements also encloses rich
information about articulateness and emotion.Automatic Speaker Recognition technology
can be defined as it is a task by which recognize speakers from their speech/voice. Speaker
Recognition has covered several speaker specific tasks, the task can be categories as text
dependent and text independent. Speaker recognition can be dividing in some specific task
such as speaker verification, speaker identification, speaker clustering, speaker
segmentation, speaker diarization and speaker detection etc.In this paper discussed about
the prosodic features for feature extraction from a speech signal. In term of speaker
recognition prosodic compute many features as duration, pitch, intensity, speech rate, tone,
stress etc.