› Forums › Automatic speech recognition › Dynamic Time Warping (DTW) › confusion about lexically contrastive
- This topic has 5 replies, 2 voices, and was last updated 1 year, 8 months ago by Simon.
-
AuthorPosts
-
-
November 28, 2022 at 16:00 #16559
Hi, i am confused about the why in engish, f0 is not lexically contrastive , cannot help to distinguish one word from another?
H searched speech perception system automatically processes fine subphonemic features, such as duration and pitch, even when they are not lexically contrastive in a language. does that mean we also will not consider the duration and pitch in the English Automatic Speech Recognition systems.
This confusio9n is based on the practise quesiton 10 in ASR.
Thank you for your help
-
November 28, 2022 at 17:24 #16562
First, remember that pitch is the perceptual correlate of F0. We can only measure F0 from a speech waveform. Pitch only exists in the mind of the listener.
When we say F0 is not lexically contrastive in ASR, we mean that it is not useful for telling two words apart. The output of ASR is the text, so we do not need to distinguish “preSENT” from “PREsent”, for example, we simply need to output the written form “present”.
Duration is lexically contrastive because there are pairs of words in the language that differ in their vowel length.
Hidden Markov Models do model duration. Can you explain how they do that?
-
December 6, 2022 at 23:21 #16632
The way that HMMs model duration is by using a set of probability distributions to represent the likelihood of different sequences of observations occurring. These probability distributions are used to calculate the likelihood of a given sequence of observations, which in turn allows the HMM to determine the most likely sequence of events that produced the observations. This allows the HMM to accurately model the duration of events, such as phonemes in speech, by taking into account the specific characteristics of the observations.
-
December 7, 2022 at 08:17 #16635
You are right that HMMs have “a set of probability distributions” but there are two different types of probability distribution in an HMM. One type is the emission probability density functions: the multivariate Gaussian in each emitting state that generates observations (MFCCs).
What is the other type?
-
December 7, 2022 at 20:56 #16650
transition probabilities ?
-
December 10, 2022 at 10:54 #16655
Correct! Can you explain how they contribute to modelling duration?
-
-
AuthorPosts
- You must be logged in to reply to this topic.