Sonic Score Saxophonics
D5: Lin Zhan, Junrui Zhao, Jordan Li
18-500 Capstone Design, Spring 2024
Electrical and Computer Engineering Department
Carnegie Mellon University
System Architecture
Product Pitch
Sonic Score Saxophonics is aimed to help aspiring saxophone players practice at home. Especially for beginners, practicing without an instructor can cause mistakes that a player may not detect. Our goal is to tackle one of the aspects of that issue, and that is note-fingering discrepancy.
Our system includes a fingering detection system installed on a real saxophone, and pitch detection algorithm, all synthesized into a webapp that helps the user identify their playing mistakes. Our fingering collection system has a 100% accuracy rate, and our pitch detection system has close to 90% accuracy rate. Feedback latency is also within 5 seconds after recording is finished.
https://course.ece.cmu.edu/~ece500/projects/s24-teamd5/
System Description
System Evaluation
Conclusions & Additional Information
Scan to access our blog site for further information
Our whole system can be divided into three subsystems: the fingering collection, audio processor and user interface. When a user plays the saxophone, its original audio will be recorded and go into the audio processor, which will first be filtered. Then the pitch detector detects the notes being played, and the rhythm detector detects the notes’ length. The output will be an integrated notes array, which will then be sent to the web app backend. The fingering data will be collected by sensors installed on the saxophone’s air doors. The controller will process and send the data to web app backend. The feedback will be generated in a few seconds.
Hardware Setup for the Saxophone
Performance of the Fingering Detector
Overall, our team achieves our goal of constructing an instructional app for saxophone beginners. Our system is able to detect notes being played by users from both input audio and sensors attached to the saxophone. We also manage to provide feedback for the users to improve their practice.
In the future, we aim to improve the accuracy of note detection for more complex songs. We hope our app can attract a broader range of users, extending its appeal beyond just saxophone beginners.
Performance of the Audico Processor
Performance of the Integrated System
Hall effect Sensors
ESP32 Microcontroller
User Interface
Output Array of Detected Notes
Detected Audio Frequency
| Accuracy | Latency |
Single Notes | 100% | 4s |
B flat Scale | 87% | 4s |
Mary Had a Little Lamb | 69% | 4s |