Category Archives: Independent Study

Audio Feature Extraction

This week I’ve continued working on the application part of my Android application, but I’ve also started to “dig deeper” (apprenticeship pattern blog post coming soon) and learn a bit more about audio processing.

I started with general concepts of feature extraction a couple of months ago now. Understanding how to use Python libraries to extract features is simple enough, but actually understanding how they work and why they work is another story. This week’s research has revealed an underlying elegance to the concept of signal processing and helped me reach a higher level of understanding and excitement for this project.

The single best explanation has been a video by Youtuber 3Blue1Brown on Fourier Series. I recommend all of his channel because he has an elegant way of describing and visualizing every topic he speaks about. He helped me understand the beauty of Calculus, and in my process of digging deeper I wound up watching his video on the uncertainty principle which was surprisingly relevant to signal processing. Understanding the specifics of the math behind signals and waves, and knowing the fact that mathematical equations are a language used to describe straightforward physical phenomena is key. This knowledge makes daunting concepts easier to break down. Seeing the same concepts used in different contexts also helps solidify them in your mind. And if you’re implementing this in code, it will make it much easier to remember the necessary logical steps required to extract a feature.

This entire tangent (and however useful, it was an unexpected tangent) started with trying to better understand the types of feature extraction that are used in speech recognition. By the way, you know you’re digging deeper when an article with an estimated read time of 11 minutes takes you a few hours to get through with all the additional research.

And universally, as far as I can tell, the first step in signal processing and feature extraction is the Fourier transform, which is simply turning a raw audio signal into separate sine and cosine signals. I say simply, but as the 3Blue1Brown video states, this seems a bit like figuring out which colors make up a mixed up can of paint. It turns out, however, that clever math makes it quite obvious which summation of sine and cosine signals make up a complex signal. I encourage you to watch the video to understand why.

The summation of cosine and sine waves is considered the frequency domain, while the original signal is in the time domain. From the resulting frequency domain, the individual signals can be normalized by taking the log magnitude of the signals and performing an inverse Fourier transform.

This is a new concept called a cepstrum, and it is one of many possible transformations you can make on a signal to begin to analyze the data. Its usefulness comes from the ability to see changes in individual waves. Additional operations can be performed to reveal new insights into patterns in a signal. Determining which if these works best is part of the process.

These individual transformations would be very interesting to implement in code. I may not get a chance to do so for this project, but the understanding of the underlying operations will help in using existing libraries.

From the blog CS@Worcester – Inquiries and Queries by James Young and used with permission of the author. All other rights reserved by the author.

Android Audio Recording and Playback, and an Animation False Start

This week, I spent a lot of time with animations in Android. And yet, I still could not get transitions to work to my liking. After reading Android’s docs, I resorted to a few different videos and tutorials, some of which seemed straightforward but were using methods of Android past.

The goal was to animate the title of a single audio file in a list, to move to the top of the screen and become a heading for details about that file. Animations are easy enough if you’re transitioning between Activities. But it appears that having a RecyclerView in a Fragment to list audio files adds some complications. I did successfully animate between the screens, but the first item in the RecyclerView’s list was animated, and it abruptly changed to the correct title when the motion ended. The issue is that this transition animation requires two elements to have a shared “transitionName”. Because I am using Card objects to display each of the audio files, only one Card can use that transitionName and be animated.

The solution is to set the name in the RecyclerView’s ViewHolder, so that when objects are bound to a specific card, they can get a unique transitionName. This can then be applied to the Fragment’s View before the animation begins. Attempts to do this caused some problems due to Android’s Lifecycle. Many a blog post have been written on this subject, but I’d like to discuss it in the near future to gain a stronger understanding.

All of this is to say: I want to efficiently getting from point A to point B by realizing which subjects and features are most important. Understanding the Android Lifecycle is clearly more important than an animation, and apparently prerequisite knowledge. And recording and playback are at the heart of the app itself. So my progress in animation is stashed in Git and ready for me to continue once I accomplish these other tasks.

Luckily, getting audio to record and play back was a much more enjoyable process. This might be due to my greater interest in the feature, but I was able to break down the problem and troubleshoot issues much easier.

I erroneously believed that my simple spike project would easily translate to my app. Android’s guide to MediaRecorder and MediaPlayer made it simple to get something quickly up and running. However, using their code directly would create a nightmare of an Activity, which neither properly separate concerns nor follow basic OOP principles. Furthermore, I needed the recording to begin immediately upon opening a new “RecordActivity”. This caused some issues with Android’s lifecycle, so I took to opportunity to explore that. The problem came from trying to start recording in onCreate(), which did not provide enough time to load the MediaRecorder into memory. The solution was to start recording on the onResume() event. However, this may be called more than once in the life of an Activity, so I simply check if the MediaRecorder is currently recording, and start recording if it isn’t.

I spent a bit too much time trying to find the recorded audio files in phone’s physical storage. It seems they do not appear, and I haven’t found a good explanation for this. Luckily, Android Studio’s Device File Explorer did reveal that the files were saved and properly recorded audio.

From there, implementing audio playback as a Service (which in Android is essentially an Activity without a UI) was rather smooth. This also allows playback to be initiated from anywhere in the app by passing the file name in a single line of code.

I have always been a function-over-form kind of guy. I made significant progress on the app this week in the function department. Hopefully the form will come in time.

From the blog CS@Worcester – Inquiries and Queries by James Young and used with permission of the author. All other rights reserved by the author.

Android Activities and Fragments: Getting Them Straight

AndroidX has brought a few changes to the Android framework, but the general architecture remains the same. Likewise, the few years since I’ve first learned Android has completely changed how I feel about Android. At this point, I am developing the basic architecture of my independent study app.

There are a lot of conflicting opinions about Activities and Fragments in the Android Developer community. A few years ago with my limited Android experience, I did not completely understand. I likely don’t completely understand now. However, I have better tools and more programming experience to see how they should be used, as well as to make my own decisions on how to use them.

At first, I found myself paralyzed with confusion on how Google wants its developers to use Activities and Fragments. As an example of how programming concepts translate well to other technologies, learning Angular helped me understand the difference. Activities are a single “thing” that a user does, and can be thought of as a web page. A Fragment should be used for a modular UI component, and function as Components do in Angular.

This isn’t a perfect analogy, as the frameworks are very different, but this is a good way to proceed when deciding how to structure your app. Google’s Introduction to App Architecture guide is a great explanation, and the most important thing to remember is to maintain a separation of concerns. In the end, Activities and Fragments aren’t a significant part of your app. They contain your app. They are something your app uses to work within the Android framework, and your business logic should be elsewhere because Android will pause or stop any Fragments or Activities it needs to if, for example, memory is running low. There is no guarantee they will maintain state unless you take additional steps to ensure it does so.

In researching opinions on how to use them, some people mentioned that they’ve seen developers decide on having a single Activity, and adding all features with Fragments. This is a tempting solution, but that might result in complex logic to control navigation. Furthermore, Fragments are meant to communicate through their parent Activity. In a large app, this would likely result in many implemented interfaces and complicated callbacks. Bloated Activities a big NO.

Likewise, others mentioned only using Activities and not adding complexity with Fragments. This seems a bit more reasonable, but restricts reusability of the Fragments in the UI. Only one Activity can run at a time. The beauty of Fragments is they can easily be dropped into a layout and reused. If they are designed to interact through their parent Activity, two Fragments can be shown at the same time on a larger tablet, even if they must be shown on different screens on a phone. Creating Activities only would mean either creating new Activities with repeated code for a tablet, or reusing the phone UI at the expense of user experience.

I’ll reiterate: separation of concerns. In Android, or any framework, understand the philosophy behind a class and component before deciding to try to simplify things. It’s likely that they were designed to prevent the problems you will run into.

From the blog CS@Worcester – Inquiries and Queries by James Young and used with permission of the author. All other rights reserved by the author.

CS-499: Independent Study Introduction

This semester, I’m building an Android app for an independent study.

The Proposal

After building a breadboard computer and beginning to understand electronics, I started to learn about audio electronics. This sparked (or reignited) a latent interest in audio processing. Working in a call center selling audio equipment is actually the reason I was motivated to return to school to study computer science, so I feel there is good reason to pursue it in my final semester.

I also began with Python, and moved to Android apps in my early programming learning process, so I’d like to refresh these skills and dig deeper. This project will serve as a constant reminder to how far I’ve come from those early struggling days.

So the app will use Python machine learning libraries to analyze user audio data and provide the user feedback based on this data. I am purposely being vague; not because I think I have the next big idea on my hands, but because I expect many changes as I struggle with the machine learning model.

Regardless of where the model winds up, this is a software development independent study. I will have a working, professional app within the next 4 months, using the technologies I have proposed.

The Motivation

Why, though? As an independent study, with an already-busy schedule, I’m going to have to set aside time each week when I work on this project, no matter what. Originally, I wanted to take Robotics this semester and I was signed up for it originally, but unfortunately there is not enough time in my schedule. On Tuesdays I’m sure I will find my mind wandering, dreaming of playing with robots instead of struggling through machine learning and Android Studio.

But that is part of my reasoning. I want to find the motivation to do things with a self-imposed deadline. These are tools I want to learn, to create and potentially sell a product. At the end of this degree, I want to be able to show a project to future employers that say, “this is what I did. Not because I had to, but because I enjoy it”. I want to be able to have users who give me unfiltered feedback. I want to fail, figure out why I failed, and eventually succeed.

Of course, I have done all of these to some extent already. But this is my following my current interests and goals.

The Progress

I have made a couple small spike projects to begin relearning Android and get started with Tensorflow. I have already built the back-end and gotten an app to communicate with it. I’ve also done basic user authentication.

When I first proposed this project, I set a schedule of features and tasks to complete. Due to other projects which used the same technologies and flashes of motivation I’ve already worked ahead a bit, but I still plan to complete each portion according to the schedule, as best as I can. The machine learning model will be concurrent work as I adjust it.

Next week, I will go into more detail on the tasks I’ve completed so far.

From the blog CS@Worcester – Inquiries and Queries by James Young and used with permission of the author. All other rights reserved by the author.