Scribd
Upload
38 views1 page

Venkata Simha Reddy

This document analyzes machine learning algorithms to convert non-audible murmur speech to normal speech with enhanced accuracy. It collects non-audible murmur speech data from speakers with various medical conditions. It then uses preprocessing, feature extraction and machine learning models like random forest, SVM, RNN, KNN and CNN for speech recognition. Based on a statistical analysis of mean accuracy across these models on over 20,000 speech samples, it finds that the random forest algorithm achieves the best accuracy of 99.86% for converting soft spoken murmur to normal speech.

Uploaded by

Caticonic
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
38 views1 page

Venkata Simha Reddy

This document analyzes machine learning algorithms to convert non-audible murmur speech to normal speech with enhanced accuracy. It collects non-audible murmur speech data from speakers with various medical conditions. It then uses preprocessing, feature extraction and machine learning models like random forest, SVM, RNN, KNN and CNN for speech recognition. Based on a statistical analysis of mean accuracy across these models on over 20,000 speech samples, it finds that the random forest algorithm achieves the best accuracy of 99.86% for converting soft spoken murmur to normal speech.

Uploaded by

Caticonic
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 1

Mr.

Desuri Venkata Simha Reddy


Guided By: Dr. Rajesh Kumar.T

Analysis of ML Algorithms with Enhanced Accuracy for


Conversion of Non-Audible Murmur to Normal Speech

INTRODUCTION

 To convert the Non-Audible Murmur to Normal Speech using Machine Learning


Techniques.
 Non-Audible Murmur (NAM) is a form of communication that allows individuals
to convey messages without speaking aloud.
 The use of this technology could bring about a significant change in communication
in situations where speaking aloud is difficult or not appropriate, for instance, in
noisy surroundings or when trying to communicate secretly.
 NAM technology is improving accessibility for individuals with disabilities,
automating call centre operations, and enabling hands-free control of devices such
as smartphones and smart speakers.

MATERIALS AND METHODS

Collection of
NAM_Speech Data Feature
Extractio
Pre-processing n
Soft Spoken
Murmur
Signal Noise removal

NAM Speaker

Recognized Audible Voice


Speech
Model ML Based Speech
Signal Recognition
Optimization

MODEL TRAINING
Audio Signal

RESULTS

 Comparison of machine learning algorithms in terms of mean accuracy. The mean accuracy of the Random Forest algorithm is
better than the SVM, RNN, KNN, and CNN.

DISCUSSION AND CONCLUSION

 The dataset is collected from the Kaggle & Timit open source website and it contains more than a lakhs of samples which is
tested from different speakers having glottal infection, throat cancer and cold sores. Twenty thousand samples were tested in the
audacity speech tools.
 Among Machine Learning algorithms, it based on the outcomes of independent T-test statistical analysis, it is determined that
the mean accuracy of two groups between
1. Recurrent Neural Network(92.11%) & Support Vector Machine(90.28%)
2. K-Nearest Neighbors(92.11%) & Recurrent Neural Network(0.939%)
3. Convolutional Neural Network(95.89%) & K-Nearest Neighbors(93.89%)
4. Random Forest(99.86%) & Convolutional Neural Network(95.89%)
 NAM Conversion is applicable in military, CBI Investigations and used by Speech impaired people for communication.
 It is concluded that the conversion of Soft Spoken Murmur to Normal Speech using Machine Learning Techniques, Random
Forest Algorithm provides best accuracy compared with the other Machine Learning Algorithms.

BIBLIOGRAPHY

Chen, Chengxin, and Pengyuan Zhang. 2022. “CTA-RNN: Channel and Temporal-Wise Attention RNN Leveraging Pre-
Trained ASR Embeddings for Speech Emotion Recognition.” Interspeech 2022. https://doi.org/10.21437/interspeech.2022-10403.

T, Rajesh Kumar, and Kumar T. Rajesh. 2021. “Enhanced Optimization in DCNN for Conversion of Non Audible Murmur to
Normal Speech Based on Dirichlet Process Mixture Feature.” Revista Gestão Inovação E Tecnologias.
https://doi.org/10.47059/revistageintec.v11i4.2239.

Babani, Denis, Tomoki Toda, Hiroshi Saruwatari, and Kiyohiro Shikano. 2011. “Acoustic Model Training for Soft Spoken
Murmur Recognition Using Transformed Normal Speech Data.” 2011 IEEE International Conference on Acoustics, Speech and
Signal Processing (ICASSP). https://doi.org/10.1109/icassp.2011.5947535

Heracleous, Panikos, and Norihiro Hagita. 2010. “Non-Audible Murmur Recognition Based on Fusion of Audio and Visual
Streams.” Interspeech 2010. https://doi.org/10.21437/interspeech.2010-717.

You might also like