Analysis of FFSR, VFSR, MFSR Techniques for Feature Extraction in Speaker Recognition: A Review
In this paper we provide the brief overview of FFSR, VFSR and
MFSR technique for speech analysis in speaker recognition
system. Following this overview we will discuss some of the
strength and weakness of current frame size and frame rate
technique. A Fixed Frame Size and Rate (FFSR) in most of stateof–
the-art speech recognition system can face some problems
such as accidentally meeting noisy frames, also assign the same
importance to each frames. As an attempt to avoid these
problems Variable Frame Size and Rate (VFSR) technique
selects optimal frame size and frame rate depending on speaking
rate to capture sudden changes in spectral information along with
time. But it increases the burden of identifying the spectral
changes in the speech. To overcome this drawback Multiple
Frame Size and Rate (MFSR) is used which provide different
feature vectors for same data which increases the performance of
speech recognition. The motivation of all these analysis is to
increase the speech recognition performance at the cost of
reducing the time and space complexity.
Keywords: Fixed Frame Size and Rate (FFSR), Variable
Frame size and Rate (VFSR), Multiple Frame Size and Rate
(MFSR)
Download Full-Text
