Himbeerdoni (Bavarian Swearword Trainer)

• 

  App UI with Himbeerdoni

• 

  Start screen

Inspiration

Bavaria and its funny swearwords, lunch talk with colleagues, the Oktoberfest

What it does

This app solves an important societal integration challenge, that is, teaching non-Bavarians how to learn and correctly pronounce the many funny swearwords of the Bavarian language.

How I built it

An iOS app records voices and submits the voice data to a backend application where audio features are extracted and compared to original audio recordings. The best match is returned to the UI.

Challenges I ran into

• Acoustic Fingerprinting (Chromaprint algorithm) seems to work well for longer music but not so well for short voice audio snippets. MFCC (Mel Frequency Cepstral Coefficients – short-term power spectrum of a sound) and comparing the features using DTW (Dynamic Time Warping – measuring similarity between two temporal sequences) has then turned out to work quite well with speech samples.
• SwiftUI: Underestimated how fundamentally the way of building iOS UIs has changed in the most recent version of iOS/Swift

Accomplishments that I'm proud of

• Feature extraction and comparison of audio files
• A cool user interface with nice graphics created by our designer

What I learned

• Audio analysis algorithms and feature extraction from audio files
• New bavarian swearwords

What's next for Bavarian Swearword Trainer

Polishing and maybe App Store Release ;)

Built With

• amazon-web-services
• python
• swift