Speech signal processing method and apparatus
AssigneeSamsung Electronics Co., Ltd.
InventorTae Gyoon KANG
Life of this patent
Estimating the term…
Abstract
A speech signal processing method and apparatus is disclosed. The speech signal processing method includes receiving an input token that is based on a speech signal, calculating first probability values respectively corresponding to candidate output tokens based on the input token, adjusting at least one of the first probability values based on a priority of each of the first probability values, and processing the speech signal based on an adjusted probability value obtained by the adjusting.
Claims23
A speech signal processing method, comprising:
- receiving an input token that is based on a speech signal;
- calculating first probability values respectively corresponding to candidate output tokens based on the input token;
- adjusting at least one of the first probability values based on a priority of each of the first probability values;
- and
- processing the speech signal based on an adjusted probability value obtained by the adjusting.
A speech signal processing apparatus, comprising:
- a receiver configured to receive an input token that is based on a speech signal;
- and
- a processor configured to
- calculate first probability values respectively corresponding to candidate output tokens based on the input token,
- adjust at least one of the first probability values based on a priority of each of the first probability values, and
- process the speech signal based on an adjusted probability value obtained by the adjusting.
A speech signal processing method, comprising:
- receiving an input token that is based on a speech signal;
- calculating first probability values respectively corresponding to first candidate output tokens based on the input token;
- calculating second probability values respectively corresponding to second candidate output tokens based on the input token;
- generating a nonverbal token and a probability corresponding to the nonverbal token based on at least one of the first probability values and the second probability values;
- and
- processing the speech signal based on the probability corresponding to the nonverbal token.
A speech signal processing method, comprising:
- receiving an input token that is based on a speech signal;
- determining, using a first neural network, first probability values respectively corresponding to first candidate output tokens based on the input token;
- determining, using a second neural network, second probability values respectively corresponding to second candidate output tokens based on the input token;
- determining whether a first probability value corresponding to a first candidate output token from among the first candidate output tokens is included in a predetermined priority,
- obtaining a weighted sum of each of the first probability values and each of the second probability values, in response to the first probability value being included in the predetermined priority;
- and
- processing the speech signal based on the weighted sum to output a text,
- wherein the first neural network is based on a language model, and the second neural network is based on a speech recognition model.
Dependencies are inferred from the claim text.