Development of Text and Speech Corpus for Designing the Multilingual Recognition System

2018 ◽  
Author(s):  
Shweta Bansal ◽  
Shyam S. Agrawal
Keyword(s):  
Recognition System ◽  
Speech Corpus

scholarly journals Database Creation and Dialect-Wise Comparative Analysis of Prosodic Features for Punjabi Language

2019 ◽  
Vol 29 (1) ◽  
pp. 1275-1282
Author(s):  
Shipra J. Arora ◽  
Rishipal Singh
Keyword(s):  
Comparative Analysis ◽  
Speech Recognition ◽  
Significant Variation ◽  
Recognition System ◽  
Speech Recognition System ◽  
Prosodic Features ◽  
Speech Corpus ◽  
Male And Female ◽  
Distinctive Features ◽  
Language Analysis

Abstract The paper represents a Punjabi corpus in the agriculture domain. There are various dialects in the Punjabi language and the main concentration is on major dialects, i.e. Majhi, Malwai and Doabi for the present study. A speech corpus of 125 isolated words is taken into consideration. These words are uttered by 100 speakers, i.e. 60 Malwi dialect speakers (30 male and 30 female), 20 Majhi dialect speakers (10 male and 10 female) and 20 Doabi dialect speakers (10 male and 10 female). Tonemes, adhak (geminated) and nasal words are selected from the corpus. Recordings have been processed through two mediums. The paper also elaborates some distinctive features of the corpus. This corpus is of quite significance for the speech recognition system. Prosodic characteristics such as intonation, rhythm and stress create a crucial impact on the speech recognition system. These characteristics vary from language to language as well as various dialects of a language. This paper portrays a comparative analysis of isolated words prosodic features of Malwi, Majhi and Doabi dialects of Punjabi language. Analysis is done using the PRAAT tool. Pitch, intensity, formant I and formant II values are extracted for toneme, adhak, nasal (bindi) and nasal (tippi) words. For all kinds of words, there is a significant variation in pitch (fundamental frequency), intensity, formant I and formant II values of male and female speakers of Malwi, Majhi and Doabi dialects. A detailed analysis has been discussed throughout this paper.

scholarly journals Compiling of Phonetic Database Structure

2021 ◽  
Vol 7 (4) ◽  
pp. 4001
Author(s):  
Maya Heydarova
Keyword(s):  
Corpus Linguistics ◽  
Field Studies ◽  
Recognition System ◽  
Speech Corpus ◽  
Linguistic Resources ◽  
Database Structure ◽  
Speech Corpora ◽  
The Voice ◽  
Source Of Information ◽  
System Today

The voice corpus of language is the essential part of the linguistic resources, and it contains the phonetic database. A phonetic database is a structured collection of software-delivered speech fragments. Nowadays, phonetic database or voice corpus became like a new element in speech technologies, and much investigation has taken place according to this event. The investigators' interest in voice corpus is related to the development of a speech recognition system. Today it is enough to experience in preparation of a phonetic database. Equipped with unique information on the preparation and usage of everyday speech corpus, the development level of speech technologies and the increasing power of computer technologies allow for the investigation of various language materials, largescale, and statistical phonetic research. These developed directions of linguistics were investigated in this article. Speech corpora are a valuable source of information for phonological research and the study of sound patterns. The study of speech corpora is in its infancy compared to other field studies in linguistics. Existing speech corpora form the part of the world's languages and do not fully represent all the dialects and speech forms by phonological aspect. The article analyses the history, structure, and importance of developing speech corpses, a branch of corpus linguistics and has developed in recent years. The article also lists the main features to be considered in the design of the speech corpus.

Improving Deep Learning based Automatic Speech Recognition for Gujarati

2022 ◽  
Vol 21 (3) ◽  
pp. 1-18
Author(s):  
Deepang Raval ◽  
Vyom Pathak ◽  
Muktan Patel ◽  
Brijesh Bhatt
Keyword(s):  
Deep Learning ◽  
Speech Recognition ◽  
Automatic Speech Recognition ◽  
Short Term Memory ◽  
Language Model ◽  
Recognition System ◽  
Processing Technique ◽  
Speech Corpus ◽  
Novel Approach ◽  
Asr System

We present a novel approach for improving the performance of an End-to-End speech recognition system for the Gujarati language. We follow a deep learning-based approach that includes Convolutional Neural Network, Bi-directional Long Short Term Memory layers, Dense layers, and Connectionist Temporal Classification as a loss function. To improve the performance of the system with the limited size of the dataset, we present a combined language model (Word-level language Model and Character-level language model)-based prefix decoding technique and Bidirectional Encoder Representations from Transformers-based post-processing technique. To gain key insights from our Automatic Speech Recognition (ASR) system, we used the inferences from the system and proposed different analysis methods. These insights help us in understanding and improving the ASR system as well as provide intuition into the language used for the ASR system. We have trained the model on the Microsoft Speech Corpus, and we observe a 5.87% decrease in Word Error Rate (WER) with respect to base-model WER.

SPEAKER IDENTIFICATION BY AGGREGATING GAUSSIAN MIXTURE MODELS (GMMs) BASED ON UNCORRELATED MFCC-DERIVED FEATURES

2014 ◽  
Vol 28 (04) ◽  
pp. 1456006 ◽  
Author(s):  
AMITA PAL ◽  
SMARAJIT BOSE ◽  
GOPAL K. BASAK ◽  
AMITAVA MUKHOPADHYAY
Keyword(s):  
Mixture Models ◽  
Speaker Recognition ◽  
Speaker Identification ◽  
Gaussian Mixture Models ◽  
Principal Component ◽  
Gaussian Mixture ◽  
Recognition System ◽  
Mel Frequency Cepstral Coefficients ◽  
Speech Corpus ◽  
Signal Process

For solving speaker identification problems, the approach proposed by Reynolds [IEEE Signal Process. Lett.2 (1995) 46–48], using Gaussian Mixture Models (GMMs) based on Mel Frequency Cepstral Coefficients (MFCCs) as features, is one of the most effective available in the literature. The use of GMMs for modeling speaker identity is motivated by the interpretation that the Gaussian components represent some general speaker-dependent spectral shapes, and also by the capability of Gaussian mixtures to model arbitrary densities. In this work, we have initially illustrated, with the help of a new bilingual speech corpus, how the well-known principal component transformation, in conjunction with the principle of classifier combination can be used to enhance the performance of the MFCC-GMM speaker recognition systems significantly. Subsequently, we have emphatically and rigorously established the same using the benchmark speech corpus NTIMIT. A significant outcome of this work is that the proposed approach has the potential to enhance the performance of any speaker recognition system based on correlated features.

scholarly journals Performance of Isolated and Continuous Digit Recognition System using Kaldi Toolkit

2019 ◽  
Vol 8 (2S2) ◽  
pp. 264-271
Keyword(s):  
System Performance ◽  
Recognition Accuracy ◽  
Language Model ◽  
Recognition Task ◽  
Recognition System ◽  
Feature Representation ◽  
Speech Corpus ◽  
Word Error Rate ◽  
Hybrid Classifier ◽  
Digit Recognition

A digit recognition system is built for recognizing the sequence of digits through 0-9. The system is experimented with speech corpus created in the room environment. The acoustic information to feature representation is achieved using PLP and MFCC features. The system initially utilized the conventional GMM-HMM framework, state of the art hybrid classifier with varied number of states to complete the speech recognition task, i.e., the system is first trained and tested using Monophone models, and system’s recognition accuracy is then evaluated using Triphone Models: Triphone1 models, which was later followed by Triphones2 models and Triphones3 Models. The Ngram Language model is used for both Monophone and Triphone training. The system performance is evaluated with the use of MFCC and PLP parameterisation techniques on Kaldi toolkit. The system performance is evaluated using metrics word error rate (WER) and Word Recognition Accuracy (WRA). The proposed system can be utilized for building speech applications

scholarly journals High level speaker specific features modeling in automatic speaker recognition system

2020 ◽  
Vol 10 (2) ◽  
pp. 1859
Author(s):  
Satyanand Singh
Keyword(s):  
Speaker Recognition ◽  
Gaussian Mixture ◽  
Recognition System ◽  
Modeling Technique ◽  
Speech Corpus ◽  
Linear Discriminant ◽  
Automatic Speaker Recognition ◽  
Wide Range ◽  
High Level ◽  
Asr System

Spoken words convey several levels of information. At the primary level, the speech conveys words or spoken messages, but at the secondary level, the speech also reveals information about the speakers. This work is based on the high-level speaker-specific features on statistical speaker modeling techniques that express the characteristic sound of the human voice. Using Hidden Markov model (HMM), Gaussian mixture model (GMM), and Linear Discriminant Analysis (LDA) models build Automatic Speaker Recognition (ASR) system that are computational inexpensive can recognize speakers regardless of what is said. The performance of the ASR system is evaluated for clear speech to a wide range of speech quality using a standard TIMIT speech corpus. The ASR efficiency of HMM, GMM, and LDA based modeling technique are 98.8%, 99.1%, and 98.6% and Equal Error Rate (EER) is 4.5%, 4.4% and 4.55% respectively. The EER improvement of GMM modeling technique based ASR systemcompared with HMM and LDA is 4.25% and 8.51% respectively.

Sign in / Sign up

Export Citation Format

Share Document