DEV Community: Nikhila K S

Harmonizing Heartbeats: A Journey into Heart Beat Sound Classification

Nikhila K S — Wed, 28 Jun 2023 09:35:03 +0000

Week 3 and 4 of my Outreach'23 internship have flown by, and I can hardly believe how quickly time has passed😌. As a second-year undergraduate student from IGDTUW, I feel incredibly grateful to have been selected as an Outreachy'23 intern.

In this blog post, I will share my experiences and progress during these two weeks of my internship. If you're interested in learning more about Outreachy, I invite you to check out my previous blogs🚀

During these weeks, our main task was to develop a machine-learning model capable of classifying heartbeat sounds. Sumaya (my co-intern) and I decided to keep our initial goal to create a model that could distinguish between "normal" and "abnormal" heartbeat sounds.

To begin, I delved deeper into the topic of sound classification with YAMNet. YAMNet is a pre-trained deep neural network developed by Google AI, specifically designed for predicting audio events from a wide range of classes. It utilizes the MobileNetV1 depthwise-separable convolution architecture, known for its efficiency and accuracy in classifying audio events. YAMNet was trained on the extensive AudioSet corpus, which consists of over 2 million audio clips sourced from YouTube videos. ( To know more checkout this repository - Link)

While exploring YAMNet, I came across an interesting project called "Heartbeat Audio Classification." This project was developed by Nittala Venkata Sai Aditya, Gandhi Disha, Saibhargav Tetali, Vishwak Venkatesh, and Soumith Reddy Palreddy as part of their Advanced Machine Learning Course in the Masters program in Business Analytics at The University of Texas at Austin. The project aimed to find the machine learning model with the highest accuracy in classifying heartbeat sound files as normal or abnormal. Since this project aligned perfectly with our weekly goal, I decided to draw inspiration from their work and train our models accordingly.

For the project, I obtained data from Peter Bentley's "Classifying Heart Sounds Challenge." This dataset included 585 labeled audio files and 247 unlabeled audio files, sourced from a clinical trial using a digital stethoscope and the iStethoscope Pro iPhone app. Since we focused on building our machine learning model, only the labeled audio files were utilized for training.

The dataset consisted of five major heart sound classes: normal, murmurs, extra heart sounds, extrasystole, and artifacts. These classes represented different characteristics of heartbeat sounds, such as distinct lub-dub patterns, whooshing or rumbling sounds, irregular rhythms, or non-heartbeat sounds. To gain insights from the data, I conducted exploratory data analysis and referred to the project team's blog, which provided valuable information about the various classes and their characteristics.

From the amplitude waveplots of the different classes, some interesting observations emerged -
Normal heart sounds exhibited a consistent distribution of amplitudes with a clear lub-dub pattern. In contrast, murmur heart sounds displayed less consistency and had additional sound waves between the lub and dub, indicating the presence of whooshing sounds. Extrasystole heart sounds had higher amplitudes and irregularities between the sound waves, indicating irregular heart rhythms. Extrahls had irregular patterns compared to normal heart sounds, with a few high-amplitude sound waves representing galloping sounds associated with certain heart conditions. Artifact class waveplots showed a wide range of different sounds, including feedback squeals, echoes, speech, music, and noise, which were unrelated to heartbeats.

Combining all the wave plots together, it became evident that extrasystole heart sounds had higher amplitudes compared to other classes, and all the heartbeat classifications exhibited irregular rhythms compared to normal heart sounds.

During the analysis, I also discovered that our dataset suffered from class imbalance, where the number of normal heartbeat sound samples far exceeded the other classes. This class imbalance posed a challenge as it could introduce bias in the model, leading to inaccurate results and an insufficient understanding of the underlying patterns distinguishing between classes.

To address this issue, we employed data augmentation techniques. With only 585 audio files for training, we decided to increase the dataset size by generating synthetic data using two popular methods: adding noise and changing pitch and speed.

Once we augmented the dataset and extracted the relevant features, we obtained 1,755 rows and 162 features. These features were essential for enabling our machine-learning models to understand the audio data. To extract the features, we utilized the librosa package in Python, a powerful tool for music and audio analysis. This package provided us with features such as Zero Crossing Rate (ZCR), Chroma, Mel Frequency Cepstral Coefficients (MFCC), RMS (Root Mean Square), and Melspectrogram, which captured different aspects of the sound waves and aided our machine learning model.

With the augmented dataset and extracted features in hand, we proceeded to train and evaluate several machine learning models on the training and test sets. The models included:

❄️Random Forest Classifier: This ensemble learning method constructs multiple decision trees and produces a majority vote for classification. We used it as our baseline model.

❄️Light Gradient Boosting Machine (LightGBM): As boosting techniques generally outperform bagging techniques, we opted for LightGBM, a faster version of Gradient Boosting Machine (GBM).

❄️CatBoost: Another boosting technique that supports categorical features and provides fast predictions.

❄️Convolutional Neural Network 1-D (CNN): CNNs excel in audio and image data analysis. We trained two CNN models, one with ReLU activation functions and another with 'tanh' activation functions to introduce nonlinearity.

All five models were trained with and without upsampling the training dataset as well with unsampled dataset. Additionally, the multi-class classification problem was converted into a binary class problem.

Upon evaluating the models' performance on the test set, we found that the random forest model and LGBMBoosting outperformed the others, achieving an accuracy of 81% and 82% respectively after class imbalance was treated.

Throughout this internship experience, I have gained valuable learnings. As a beginner in machine learning, I initially struggled to understand the codebase. However, thanks to the well-written blog and documented GitHub repository, I was able to overcome these challenges and grasp the concepts effectively.✨

As I move forward in my internship journey, I look forward to further improving our heartbeat sound classification model and exploring advanced techniques to enhance its accuracy and versatility. Additionally, I aim to create a user-friendly interface that will make our model accessible to medical professionals and researchers in the field.

In conclusion, weeks 3 and 4 of my Outreach'23 internship have been incredibly enlightening🤩. I have acquired knowledge in sound classification with YAMNet, applied it to the domain of heart beat sounds, and developed a machine learning model with promising results. I am excited to continue working on this project, facing new challenges, and witnessing its progression in the coming weeks💫.

If you're interested in knowing about my internship journey follow me on my socials, be sure to check out my previous blogs, where I have share insights into the Outreachy program and my experiences as an intern.

Stay tuned for more updates as I delve deeper into the world of machine learning and tackle new obstacles!

References

Medium Blog of the project I referred - Link
My GitHub Repository - Link

Diving into Machine Learning🤖

Nikhila K S — Thu, 15 Jun 2023 12:18:07 +0000

Welcome back to my Outreachy blog series, where I share my progress and experiences during my Outreachy internship.

I am Nikhila K S an undergraduate student from IGDTUW India (Socials Link Tree). If you want to know more about Outreachy check out my previous blogs🌸.

Now back to week 2😄
My co-intern Sumaya and I embarked on an exciting adventure into the world of machine learning. Join us as we explore different tools, datasets, and learning opportunities that have shaped our understanding and enthusiasm for this fascinating field.

⚡Exploring Tools: Google Teachable Machine, Edge Impulse, and Google Colab
During this week, Sumaya, who had prior experience in machine learning, and I, an absolute beginner, ventured into the vast realm of machine learning. We began our exploration by investigating various tools like Google Teachable Machine, Edge Impulse, and Google Colab, all of which held promise for developing our machine-learning model. After careful consideration, we concluded that Google Colab provided the flexibility and control we needed for our project. Its extensive library and resources made it the ideal platform for building our machine-learning model.

⚡Discovering Valuable Datasets on Kaggle
To effectively train our machine learning model, we set out to find suitable datasets. Our journey led us to Kaggle, a renowned platform for data science enthusiasts. Here, we discovered a plethora of valuable datasets that aligned perfectly with our project requirements. These datasets would serve as the foundation for training and evaluating our model accurately.

Some of our References -

⚡Choosing the Optimal Cloud Storage Solution: Google Drive
With an abundance of data at our disposal, we needed a reliable cloud storage solution. We explored various options, such as Google Drive, Mega Account, and Firestore, to determine the most suitable choice. After careful deliberation, we unanimously decided to move forward with Google Drive. Its seamless integration with other Google services and user-friendly interface made it the ideal platform for organizing and managing our project files.

⚡Gaining Insights from Similar Projects-
To enhance our understanding of machine learning concepts and explore different approaches, we dedicated time to studying repositories and documentation related to similar projects. This deep dive allowed us to glean valuable insights and identify potential strategies to successfully complete our project. By building upon the knowledge and experiences shared by others, we could shape our own unique approach.

⚡Research Findings
How to train the heart or lung's sound data to detect anomalies.

Digital Stethoscope AI App
This project was shared by mentor @ruqaiya
Uses ML libraries from Google Tensorflow to train sound data. It looks out for anomalies in sound leveraging the features of the sound. After which it then classifies the sounds into various types of heart/lung conditions(respiratory tract infection e.g. COVID-19 or Pneumonia). Among the Classification labels is Normal.
Bird Sound Identifier on Native Android Project
This project identified bird sound using the frequency of the sound. It deployed the trained data in tflite data type to Native Android for testing. It also has a much better version of where it began to look out for the types of birds-🔗[This video combines the bird sound identifier with a further analysis of bird type, it has a corresponding video made by Tensorflow on how to achieve it].
Sound classification with YAMNet
This is a Tensorflow project for Classifying sound data. A couple of projects we looked into used this implementation. It follows the most basic approach to Classifying sound data leveraging the TensorFlow library.

⚡Mentor Meeting and Positive Reviews🚀
Over the weekend, we had the opportunity to present our progress during a mentor meeting. Through a detailed presentation, we showcased our achievements and received positive reviews from our mentors. This feedback served as a motivation booster, fueling our determination to excel in our project and exceed expectations.

⚡Embracing the Learning Journey in Machine Learning
As a beginner in machine learning, I found this week's exploration to be both exciting and enjoyable. Although learning machine learning in a week is a formidable task, I firmly believe that continuous learning while actively working on our project tasks will significantly enhance my understanding and expertise in this field. To deepen my knowledge, I delved into Sound classification with YAMNet and compiled detailed notes, solidifying my grasp of the topic. (File link)

⚡A new habit🤗...
I embraced a new habit of documenting my learning journey on GitHub, recognizing its advantages as a centralized hub. This practice not only enables me to quickly revisit and reinforce my understanding of concepts but also eliminates the need for excessive computer storage. GitHub has now become my reliable one-stop destination to reflect on and review my progress whenever I desire. If you are starting with a new technology I would suggest you to do the same🚀.
Wanna check out my repo? here you go, I will keep updating the repo as I learn - Unravel_ML

Week 2 of my Outreachy internship immersed me in the captivating world of machine learning, leaving an indelible mark on my journey⚡. As I conclude this blog, I carry the spirit of continuous improvement, driving me to surpass achievements each week.
With an insatiable thirst for knowledge, I eagerly embrace the upcoming weeks, equipped with newfound expertise and an unwavering determination to thrive in my Outreachy project. Stay connected as I navigate this exhilarating path, where I will unravel the endless possibilities of machine learning, making a lasting impact through our dedication, passion, and unyielding pursuit of success.

Have a nice day🌸

My Outreachy Journey: Embracing Core Values and Exploring the Open Source World

Nikhila K S — Tue, 06 Jun 2023 09:11:25 +0000

Hello, dear readers!

I am thrilled to be sharing my first blog post about my incredible journey till now as an intern in Outreachy. This opportunity has already proven to be life-changing for me, and I can't wait to take you through my experiences and lessons learned along the way.

Allow me to introduce myself. I am a second-year engineering student, and being the first in my family to pursue this field, I often find myself overwhelmed by the vast array of technologies and possibilities that lie ahead.

My core values revolve around three key principles: a strong commitment to continuous learning and going the extra mile to accomplish tasks, an enthusiastic drive to explore new frontiers, and a sincere dedication to being an attentive listener. These values form the foundation of my character and shape my approach to life and work.

First and foremost, my unwavering willingness to learn propels me to continually expand my knowledge and skills. I embrace challenges and strive to exceed expectations, consistently pushing myself to achieve more.

Secondly, my inherent enthusiasm drives me to explore new realms and embrace novel experiences. I thrive on the excitement of discovering uncharted territories, pushing boundaries, and seeking fresh perspectives.

Lastly, being a good listener allows me to connect with others on a deeper level. By actively listening, understanding, and empathizing with different viewpoints, I foster meaningful relationships and create an inclusive environment where everyone feels valued.

These core values act as my guiding principles, inspiring me to pursue personal and professional growth while maintaining a strong sense of integrity and empathy. By embodying these values, I aim to make a positive impact and continually strive for excellence in all aspects of life.

It was less than a year ago that I joined LinkedIn, and witnessing the achievements of people around the world ignited a fire🔥 within me, driving me to work harder and aim higher🚀.

Last October, I stumbled upon the world of Open Source, and it was truly an eye-opening experience.🤩 This newfound knowledge opened doors to a multitude of opportunities for me to learn and grow. Conversations with my college seniors further enlightened me about remarkable open-source programs such as Outreachy, GSOC, and GSOD. Intrigued by the positive impact and immense potential of open source, I immersed myself in articles and blogs that delved into its causes and advantages. It was during this exploration that I felt an overwhelming desire to give it a shot and contribute in my own way😊.

From December onwards, I eagerly awaited the opening of the Outreachy application. Although I was a complete beginner in different technologies and open source, a sense of determination overshadowed any fears or anxieties I had. I firmly believed that giving my best, regardless of the outcome, was of paramount importance.

The initial application process involved answering eligibility questions and completing a set of essay questions that allowed me to express my passion and drive for this program. I completed and submitted my application within a week of its opening.

In February, the results were announced, and to my utmost delight, I was selected in the initial round.

Now, the real hustle began – the Contribution Period. Outreachy released a list of companies and their projects participating in the Outreachy '23 program. Being relatively new to Flutter, a technology I had been exploring for a few months, I searched for projects related to Flutter and was fortunate to stumble upon a captivating project called "Improve-a-digital-Stethoscope-app" by Mboalab Cameroon. Excitement surged through me as I joined their Slack and WhatsApp groups, where the mentors and fellow contributors warmly welcomed me and provided invaluable guidance.

My mentors, Jafsia Elisee and Ruqaiya Sattar, explained the tasks that needed to be completed during the Contribution Period. Despite my limited familiarity with Flutter, I was determined not to give up halfway. I immersed myself in documentation, articles, and studied similar project codes to gain a better understanding. After much effort, I proudly submitted my first pull request (PR), and receiving positive feedback from my mentor filled me with joy and motivation. I continued to strive for improvement with each subsequent PR, constantly working to enhance the quality of my contributions. Before I knew it, a month had flown by, and I was filled with a deep sense of satisfaction as several of my PRs were merged into the project.

As the time to apply for an internship position approached, I had to submit a detailed timeline along with essays that described why I wished to intern with Mboalab, Outreachy, and GitHub repositories showcasing my skills relevant to the project, as well as the PRs I had created during the Contribution Period. With a mixture of excitement, anticipation, and nervousness, I submitted my application and anxiously awaited the final results.

On May 4, 2023, a wave of euphoria swept over me as the results were announced, and I received the incredible news that I had been selected as an intern. It felt as if I were on cloud nine, overwhelmed with happiness and a deep sense of accomplishment. The journey I had embarked upon with Outreachy exceeded my expectations, leaving me with cherished memories that will forever hold a special place in my heart.

I am immensely grateful to my mentor, Ruqaiya Sattar, whose unwavering support and prompt responses to my queries played a pivotal role in my success. Her guidance and encouragement gave me the confidence to navigate the challenges and explore new horizons during this transformative experience.

In my upcoming blog posts, I will delve into the fascinating journey that awaits me as an intern at Outreachy. I will share the remarkable insights, lessons, and triumphs I encounter along the way, providing a glimpse into the world of open source and the impactful work being done by dedicated individuals. Stay tuned as I embark on this amazing adventure, and together we will uncover the intricacies and joys of being an intern in Outreachy.

💠 Outreachy Website - https://www.outreachy.org/

💠 My Socials - 🔗LinkTree

Paving the Way: Understanding the Outreachy Application Process

Nikhila K S — Tue, 06 Jun 2023 07:19:46 +0000

💠Content of this blog :

What is Outreachy?
Eligibility
Timeline
How to apply -
Initial application
Contribution period
Internship period

💠 What is Outreachy?

Outreachy, previously known as the Free and Open Source Software Outreach Program for Women, is a pioneering initiative that seeks to bridge the gap of underrepresentation in free and open-source software projects. The program orchestrates three-month paid internships, specifically designed for individuals who belong to communities traditionally marginalised within these projects. Originally organised by The GNOME Project and the GNOME Foundation, it is now overseen by the Software Freedom Conservancy.

By providing these paid remote internships, Outreachy extends a vital opportunity for participants to immerse themselves in the world of free and open-source software, while receiving financial support for their contributions. The program's overarching goal is to cultivate a more inclusive and diverse tech landscape. Through hands-on experience and collaboration with established projects, Outreachy empowers individuals from underrepresented backgrounds to unlock their potential and make meaningful contributions to the open-source community. By fostering a culture of inclusivity, the program actively challenges and dismantles barriers that hinder diversity within the tech industry.

💠 Eligibility

Outreachy is open to applicants around the world.
They invite people to apply who face systemic bias or discrimination in the technology industry of their country.
The following are some criteria as stated on the official website -
1. General eligibility
You must be 18 years of age or older by May 29, 2023
You must be available for a full-time internship. Outreachy interns work 30 hours per week. The internship runs from May 29, 2023 to Aug. 25, 2023.
2. Past internships
You are welcome to apply to Outreachy multiple times. However, you can only be accepted as an Outreachy intern once.
You must not be a past Outreachy intern.
You must not be a past Outreach Program for Women intern.
You must not be a past Google Summer of Code intern. All Google Summer of Code interns are ineligible for Outreachy. This includes people who did not successfully finish their Google Summer of Code internship.
3. Current or future internships
The Outreachy internship runs from May 29, 2023 to Aug. 25, 2023.
You must not have another internship during the Outreachy internship period. This includes unpaid internships.
Applicants are required to list their current internships on their initial application. We understand you may be applying to many jobs. If you receive a job or internship offer, please notify Outreachy organizers immediately.
4. Rules for people with jobs
The Outreachy internship runs from May 29, 2023 to Aug. 25, 2023.
You must not have a full-time job during the Outreachy internship.
You must not have a full-time contracting position during the Outreachy internship period.
You must not be on a leave of absence from a full-time job during the Outreachy internship.
If you are willing to quit your full-time job, you are welcome to apply to Outreachy. If you cannot quit your full-time job, you are not eligible for Outreachy.
If you have a part-time job, you are welcome to apply to Outreachy. Part-time jobs must be approved by Outreachy organizers.
Applicants are required to list their current jobs on their initial application. We understand you may be applying to many jobs. If you receive a job or internship offer, please notify Outreachy organizers immediately.
5. Rules for people who are not students
People who are not students are welcome to apply to Outreachy.
Outreachy has two internship cohorts: May to August, and December to March. If you are not a student, you may apply to either internship cohort.
6. Rules for students
Both students and people who are not students are welcome to apply to Outreachy.
University students must have 42 consecutive days free from school and exams during the internship period.
Students must apply to the correct internship cohort (see rules below).
Outreachy internships run twice a year, May to August and December to March. We have some rules around which internship round you can apply to:
If you are a student of a university in the Northern Hemisphere, you will only be eligible for the May to August internship cohort. Students in India are considered to be in the northern hemisphere, regardless of where their university is located.
If you are a student of a university in the Southern Hemisphere, you will only be eligible for the December to March internship cohort.
Otherwise, if your university is near the equator, you may apply to any internship cohort. We will review university term schedules on a case-by-case basis.
If you are completing your last term for your degree, you may be eligible for either internship cohort, regardless of what hemisphere you are in. To be eligible for the May 2023 internships, the exams for your last school term must end before July 10, 2023. If the school academic calendar continues past that date, graduating students are not eligible.
When determining student eligibility, Outreachy looks at the school's academic calendar dates. We do not consider the individual student's course load. We do not consider special arrangements students have made with their university. Students who plan on taking a school term off to pursue an internship are not eligible for Outreachy. We cannot accept letters from universities about pausing classes or joining classes at a later date.
7. Rules for students on visas
Your visa must allow you to work 30 hours per week. If you cannot work 30 hours per week, you are not eligible for Outreachy.
If you are on a student visa in the United States of America, you might have limited dates when you can work 30 hours a week. We will work with you to shift your internship dates by up to five weeks. However, we cannot accommodate shortening the 13 week internship.
If you are a student on an F-1 visa, you may need to apply for CPT with your university. Outreachy organizers can provide you documentation for your CPT application once you are selected as an intern.

💠Timeline-

For May- August internship period

Intern applications are accepted from mid-January to early February
Internship project list is finalized in early March 2023
Mentors can sign up starting in January
Contribution period starts from March till April
Internships run late May 2024 to late August 2024

For December-March internship period

Intern applications are accepted from early August to late August 2023
Internship project list will be finalized in late September 2023
Mentors can sign up starting in early August 2023
Internships run early December 2023 to early March 2024

💠How to apply -

🔶Initial Application -
The initial application round for the Outreachy program involves a series of questions designed to assess your eligibility. These questions are accompanied by four essay prompts. Starting from 2023, Outreachy has adopted a first-come, first-served approach, underscoring the importance of submitting your application as early as possible.
The essay questions, each with a character limit of 1000, are as follows:

Are you a member of an underrepresented group in the technology industry of your country? How do you experience underrepresentation?
If you were to apply for a job in the technology industry of your country, what systemic biases or discrimination do you anticipate encountering?
Does your learning environment have a limited number of individuals who share your identity or background? Please provide specific details.
What systemic biases or discrimination have you encountered while developing your skills? For more information, including additional details, please visit the official website - Applicant Guide|Outreachy Documentation If you are selected in the application round you will receive a mail inviting you for contribution period.

🔶Contribution Period -
Once the contribution period commences, Outreachy publishes a comprehensive list of participating companies and their projects for the year. As contributors, we have the freedom to explore these projects and choose the one we are interested in contributing. Each project description includes the specific tasks that contributors are expected to complete.
To make the most of the contribution period, consider the following tips:

Embrace your confidence and acknowledge that you possess unique strengths and abilities, just like everyone else.
Thoroughly familiarize yourself with the project you select to ensure a deeper understanding of its requirements and objectives.
When submitting pull requests (PRs), ensure that they are well-defined and accompanied by clear comments, outlining the work you have accomplished. This practice helps provide transparency and clarity about your contributions.
Pro Tip: Stay engaged in the community communication channels by actively participating in discussions, assisting others with their queries, and seeking help when you encounter obstacles. By doing so, you will not only find enjoyment in the experience but also have the opportunity to connect with individuals from diverse backgrounds and learn about their unique skills, which can be truly fascinating.

By following these suggestions, you can maximize your engagement during the contribution period and make meaningful contributions to the chosen project.

🔶Results

Once the contribution period concludes, we are required to submit an application for the internship we desire. This application includes essential details and documents such as a proposed timeline outlining the work we intend to accomplish during the internship, a summary of our contributions during the contribution period, and a showcase of relevant projects that demonstrate our possession of the required skills.

Approximately a month later, Outreachy announces the outcomes. In August 2023, a total of 63 interns were selected for the program.

I take immense pride in being chosen as one of the interns for the August 2023 cohort. It was an incredibly overwhelming moment for me. Throughout my Outreachy journey, I will be documenting my progress in biweekly blog posts. Please feel free to visit my profile to access and read about my experiences.

Outreachy Website - https://www.outreachy.org/

My Socials - 🔗LinkTree