Estimating BMI Through Speech Signals: A Computational Approach

Abstract

Obesity is a serious health problem worldwide because of the danger factors associated with diseases which cause permanent psychological effect. To classify normal weight, overweight, underweight and obesity, body mass index (BMI) is the most recognized and extensively used measurement. BMI measurement has its limits in some cases like overestimate in athletes, and underestimates in elderly. Thus, the project, reports the estimation of BMI (body mass index) status via speech signals. The deployment of the system is very computationally lightweight.

It uses very short audio samples; 5sec (approx.).The audio of the end user is not recorded thus ensuring privacy.

Introduction

Obesity is a precursor to numerous chronic diseases, emphasizing the need for effective measurement tools like BMI. This project introduces a novel method to estimate BMI using voice samples, aiming to simplify the measurement process while ensuring user privacy.

Need for the Project

Nowadays, rapid increasing numbers of obese people are becoming worldwide health concern. The victim across board these include adults, adolescents, children and both men and women vulnerable to this menace.

Get quality help now

Writer Lyla

Verified writer

Proficient in: Science

5 (876)

“ Have been using her for a while and please believe when I tell you, she never fail. Thanks Writer Lyla you are indeed awesome ”

+84 relevant experts are online

Hire writer

In scientific sense, the obesity refers to excess adipose tissue caused by genetic determinants, excessive taking, insufficient physical exercise and movement. Also due to some an inappropriate lifestyle has been reported as part of this problem. Obesity has a straight connection with physical health and psychological health and is a potential risk factor for many illnesses, including stroke, ischemic heart disease , cardiovascular diseases, cancer, and diabetes. The most obese are prone to fatal disease.

Regardless of the economic development, many countries are suffering from obesity epidemic; Obesity has become a cause of death, According to a WHO report in 2010, each year more than 2 billion people died because of obesity and 2,3 billion will be overweight and 700 million obese in 2015.

The reductions in the rate of obesity in developed societies are incomplete so far due to the integration of physiological, behavioral, social and cultural rights of the population, it is now priority of every county to participate in controlling obesity because most obese are susceptible to die young.

Literature Survey

Chawki Berkai [1] in his paper he review’s significant research works on prediction of BMI status using speech signals. The three important issues in previous projects were: the important design criteria of databases, the features used to characterize height and weight and the classification techniques used. The average classification accuracy to determine BMI status is less than 75% in most of the proposed techniques and still needs significant improvement.

Berkai, Chawki [2] in his paper demonstrates a comparison of two speech feature extraction methods for the classification that are KNN and PNN of BMI status. MFCCs, LPCs, LPCCs, WLPCCs cepstral features were extracted from the speech signals. PNN gives accuracy of 87.90±0.47 using MFCCs and 83.06±0.61 using LPCCs and KNN gives accuracy of 87.74 ±0.46 using MFCCs and 82.48± 0.47 using LPCCs. Hence the study has demonstrated that PNN outperformed KNN and gives the best average classification accuracy to estimate/classify the BMI Status The study has demonstrated that PNN outperformed kNN and gives the best average classification accuracy to estimate/classity the BMI Status.

Objectives

• To calculate BMI (Body Mass Index) accurately over voice.
• It is portable, and efficient.

Methodology

Discourse analysis aims to automatically identify the various information of a human such as language, emotion, age, gender, height etc, with extracting appropriate features and take suitable methods. Figure 1 suggests components of the speech system.

Implementation

Code for Sound Recording:

import numpy as np import os

from scipy.io.wavfile import write duration = 2

fs = 44100

if not os.path.isdir('Samples'): os.mkdir('Samples')a

commands = ['A', 'E', 'I', 'O', 'U']

for i in commands:

if not os.path.isdir('Samples/' + i): os.mkdir('Samples/' + i)

for cmd in commands: print('Now say: ' + cmd) time.sleep(1)

for i in range(20): print('Record ' + str(i + 1))

record = sd.rec(int(duration*fs), samplerate = fs, channels=2) sd.wait()

path = 'Samples/' + cmd + '/' + cmd + str(i) + '_1.wav' write(path, fs, scaled)

time.sleep(2)

print('Thank You So Much For Helping !!! ;)')

Proposal for SEM VII

• Removal of noise from data.
• Training the model.
• Processing of data.

Conclusion

This project demonstrates a Python-based approach to record sound for BMI estimation, planning to process and train a model with a large dataset of voice samples. This innovative method promises a new direction in health monitoring, leveraging technology to address the obesity epidemic.

References

1. Chawki Berkai, M. Hariharan. "Prediction of Body Mass Index using Speech Signals: A Review."
2. Chawki Berkaia, M. Hariharana, Sazali Yaacobb, Mohd Iqbal Omara. "Estimation of BMI Status via Speech Signals using Short-term Cepstral Features."