Performance of forced-alignment algorithms on children’s speech

2020 ◽  
Author(s):  
Tristan Mahr ◽  
Visar Berisha ◽  
Kan Kawabata ◽  
Julie Liss ◽  
Katherine Hustad
Keyword(s):  
Speech Recognition ◽  
Gold Standard ◽  
Alignment Accuracy ◽  
Speech Sample ◽  
Older Children ◽  
Adaptive Training ◽  
Alignment Algorithms ◽  
Child Speech ◽  
Speech Recognition Engine ◽  
Children's Speech

Aim. We compared the performance of five forced-alignment algorithms on a corpus of child speech.Method. The child speech sample included 42 children between 3 and 6 years of age. The corpus was force-aligned using the Montreal Forced Aligner with and without speaker adaptive training, triphone alignment from the Kaldi speech recognition engine, the Prosodylab Aligner, and the Penn Phonetics Lab Forced Aligner. The sample was also manually aligned to create gold-standard alignments. We evaluated alignment algorithms in terms of accuracy (whether the interval covers the midpoint of the manual alignment) and difference in phone-onset times between the automatic and manual intervals.Results. The Montreal Forced Aligner with speaker adaptive training showed the highest accuracy and smallest timing differences. Vowels were consistently the most accurately aligned class of sounds across all the aligners, and alignment accuracy increased with age for fricative sounds across the aligners too. Interpretation. The best-performing aligner fell just short of human-level reliability for forced alignment. Researchers can use forced alignment with child speech for certain classes of sounds (vowels, fricatives for older children), especially as part of a semi-automated workflow where alignments are later inspected for gross errors.

scholarly journals Performance of Forced-Alignment Algorithms on Children's Speech

2021 ◽  
pp. 1-10
Author(s):  
Tristan J. Mahr ◽  
Visar Berisha ◽  
Kan Kawabata ◽  
Julie Liss ◽  
Katherine C. Hustad
Keyword(s):  
Gold Standard ◽  
Acoustic Measurement ◽  
Manual Segmentation ◽  
Speech Sample ◽  
Older Children ◽  
Adaptive Training ◽  
Alignment Algorithms ◽  
Child Speech ◽  
Speech Recognition Engine ◽  
Children's Speech

Purpose Acoustic measurement of speech sounds requires first segmenting the speech signal into relevant units (words, phones, etc.). Manual segmentation is cumbersome and time consuming. Forced-alignment algorithms automate this process by aligning a transcript and a speech sample. We compared the phoneme-level alignment performance of five available forced-alignment algorithms on a corpus of child speech. Our goal was to document aligner performance for child speech researchers. Method The child speech sample included 42 children between 3 and 6 years of age. The corpus was force-aligned using the Montreal Forced Aligner with and without speaker adaptive training, triphone alignment from the Kaldi speech recognition engine, the Prosodylab-Aligner, and the Penn Phonetics Lab Forced Aligner. The sample was also manually aligned to create gold-standard alignments. We evaluated alignment algorithms in terms of accuracy (whether the interval covers the midpoint of the manual alignment) and difference in phone-onset times between the automatic and manual intervals. Results The Montreal Forced Aligner with speaker adaptive training showed the highest accuracy and smallest timing differences. Vowels were consistently the most accurately aligned class of sounds across all the aligners, and alignment accuracy increased with age for fricative sounds across the aligners too. Conclusion The best-performing aligner fell just short of human-level reliability for forced alignment. Researchers can use forced alignment with child speech for certain classes of sounds (vowels, fricatives for older children), especially as part of a semi-automated workflow where alignments are later inspected for gross errors. Supplemental Material https://doi.org/10.23641/asha.14167058

scholarly journals Examining Factors Influencing the Viability of Automatic Acoustic Analysis of Child Speech

2018 ◽  
Vol 61 (10) ◽  
pp. 2487-2501 ◽  
Author(s):  
Thea Knowles ◽  
Meghan Clayards ◽  
Morgan Sonderegger
Keyword(s):  
Young Children ◽  
Acoustic Analysis ◽  
Center Of Gravity ◽  
Training Data ◽  
Alignment Accuracy ◽  
Older Children ◽  
Acoustic Characteristics ◽  
Phonetic Transcription ◽  
Very Young Children ◽  
Child Speech

Purpose Heterogeneous child speech was force-aligned to investigate whether (a) manipulating specific parameters could improve alignment accuracy and (b) forced alignment could be used to replicate published results on acoustic characteristics of /s/ production by children. Method In Part 1, child speech from 2 corpora was force-aligned with a trainable aligner (Prosodylab-Aligner) under different conditions that systematically manipulated input training data and the type of transcription used. Alignment accuracy was determined by comparing hand and automatic alignments as to how often they overlapped (%-Match) and absolute differences in duration and boundary placements. Using mixed-effects regression, accuracy was modeled as a function of alignment conditions, as well as segment and child age. In Part 2, forced alignments derived from a subset of the alignment conditions in Part 1 were used to extract spectral center of gravity of /s/ productions from young children. These findings were compared to published results that used manual alignments of the same data. Results Overall, the results of Part 1 demonstrated that using training data more similar to the data to be aligned as well as phonetic transcription led to improvements in alignment accuracy. Speech from older children was aligned more accurately than younger children. In Part 2, /s/ center of gravity extracted from force-aligned segments was found to diverge in the speech of male and female children, replicating the pattern found in previous work using manually aligned segments. This was true even for the least accurate forced alignment method. Conclusions Alignment accuracy of child speech can be improved by using more specific training and transcription. However, poor alignment accuracy was not found to impede acoustic analysis of /s/ produced by even very young children. Thus, forced alignment presents a useful tool for the analysis of child speech. Supplemental Material https://doi.org/10.23641/asha.7070105

scholarly journals Usage of Prosody Modification and Acoustic Adaptation for Robust Automatic Speech Recognition (ASR) System

2021 ◽  
Vol 35 (3) ◽  
pp. 235-242
Author(s):  
Vivek Bhardwaj ◽  
Vinay Kukreja ◽  
Amitoj Singh
Keyword(s):  
Speech Recognition ◽  
Automatic Speech Recognition ◽  
Research Work ◽  
Recognition Rate ◽  
Speaking Rate ◽  
Word Error Rate ◽  
Adaptive Training ◽  
Prosody Modification ◽  
Children's Speech ◽  
Asr System

Most of the automatic speech recognition (ASR) systems are trained using adult speech due to the less availability of the children's speech dataset. The speech recognition rate of such systems is very less when tested using the children's speech, due to the presence of the inter-speaker acoustic variabilities between the adults and children's speech. These inter-speaker acoustic variabilities are mainly because of the higher pitch and lower speaking rate of the children. Thus, the main objective of the research work is to increase the speech recognition rate of the Punjabi-ASR system by reducing these inter-speaker acoustic variabilities with the help of prosody modification and speaker adaptive training. The pitch period and duration (speaking rate) of the speech signal can be altered with prosody modification without influencing the naturalness, message of the signal and helps to overcome the acoustic variations present in the adult's and children's speech. The developed Punjabi-ASR system is trained with the help of adult speech and prosody-modified adult speech. This prosody modified speech overcomes the massive need for children's speech for training the ASR system and improves the recognition rate. Results show that prosody modification and speaker adaptive training helps to minimize the word error rate (WER) of the Punjabi-ASR system to 8.79% when tested using children's speech.

scholarly journals Realization of speech in design intent of geometric modelling

2017 ◽  
Vol 7 (1.3) ◽  
pp. 121
Author(s):  
Sreeja B P ◽  
Amrutha K G ◽  
Jeni Benedicta J ◽  
Kalaiselvi V ◽  
Ranjani R
Keyword(s):  
Speech Recognition ◽  
Geometric Modeling ◽  
Geometric Modelling ◽  
Design Intent ◽  
Voice Input ◽  
Interactive Mode ◽  
Modeling Software ◽  
Speech Recognition Engine ◽  
The Voice ◽  
Geometric Designs

The conventional interactive mode is especially used for geometric modeling software. This paper describes, a voice-assisted geometric modeling mechanism to improve the performance of modeling, speech recognition technology is used to design this model. This model states that after receiving the voice command, the system uses the speech recognition engine to identify the voice commands, then the voice commands identified are parsed and processed to generate the geometric design based on the users voice input dimensions, The outcome of the system is capable of generating the geometric designs to the user via speech recognition. This work also focuses on receiving the feedback from the users and customized the model based on the feedback.

scholarly journals Vargas: heuristic-free alignment for assessing linear and graph read aligners

2019 ◽  
Author(s):  
Charlotte A. Darby ◽  
Ravi Gaddipati ◽  
Michael C. Schatz ◽  
Ben Langmead
Keyword(s):  
Gold Standard ◽  
Source Code ◽  
Alignment Accuracy ◽  
Local Alignment ◽  
Maximum Speed ◽  
Command Line ◽  
Scoring Functions ◽  
Large Numbers ◽  
Computationally Intensive ◽  
Optimal Alignments

AbstractRead alignment is central to many aspects of modern genomics. Most aligners use heuristics to accelerate processing, but these heuristics can fail to find the optimal alignments of reads. Alignment accuracy is typically measured through simulated reads; however, the simulated location may not be the (only) location with the optimal alignment score. Vargas implements a heuristic-free algorithm guaranteed to find the highest-scoring alignment for real sequencing reads to a linear or graph genome. With semiglobal and local alignment modes and affine gap and quality-scaled mismatch penalties, it can implement the scoring functions of commonly used aligners to calculate optimal alignments. While this is computationally intensive, Vargas uses multi-core parallelization and vectorized (SIMD) instructions to make it practical to optimally align large numbers of reads, achieving a maximum speed of 456 billion cell updates per second. We demonstrate how these “gold standard” Vargas alignments can be used to improve heuristic alignment accuracy by optimizing command-line parameters in Bowtie 2, BWA-MEM, and vg to align more reads correctly. Source code implemented in C++ and compiled binary releases are available at https://github.com/langmead-lab/vargas under the MIT license.

scholarly journals Prosodic Feature-Based Discriminatively Trained Low Resource Speech Recognition System

2022 ◽  
Vol 14 (2) ◽  
pp. 614
Author(s):  
Taniya Hasija ◽  
Virender Kadyan ◽  
Kalpna Guleria ◽  
Abdullah Alharbi ◽  
Hashem Alyami ◽  
...  
Keyword(s):  
Speech Recognition ◽  
Mutual Information ◽  
Data Augmentation ◽  
Recognition System ◽  
Speech Recognition System ◽  
Prosodic Features ◽  
Prosodic Feature ◽  
Feature Based ◽  
Maximum Mutual Information ◽  
Children's Speech

Speech recognition has been an active field of research in the last few decades since it facilitates better human–computer interaction. Native language automatic speech recognition (ASR) systems are still underdeveloped. Punjabi ASR systems are in their infancy stage because most research has been conducted only on adult speech systems; however, less work has been performed on Punjabi children’s ASR systems. This research aimed to build a prosodic feature-based automatic children speech recognition system using discriminative modeling techniques. The corpus of Punjabi children’s speech has various runtime challenges, such as acoustic variations with varying speakers’ ages. Efforts were made to implement out-domain data augmentation to overcome such issues using Tacotron-based text to a speech synthesizer. The prosodic features were extracted from Punjabi children’s speech corpus, then particular prosodic features were coupled with Mel Frequency Cepstral Coefficient (MFCC) features before being submitted to an ASR framework. The system modeling process investigated various approaches, which included Maximum Mutual Information (MMI), Boosted Maximum Mutual Information (bMMI), and feature-based Maximum Mutual Information (fMMI). The out-domain data augmentation was performed to enhance the corpus. After that, prosodic features were also extracted from the extended corpus, and experiments were conducted on both individual and integrated prosodic-based acoustic features. It was observed that the fMMI technique exhibited 20% to 25% relative improvement in word error rate compared with MMI and bMMI techniques. Further, it was enhanced using an augmented dataset and hybrid front-end features (MFCC + POV + Fo + Voice quality) with a relative improvement of 13% compared with the earlier baseline system.

Sign in / Sign up

Export Citation Format

Share Document