Portfolio

Research

Timbre Perception, Representation, and its Neuroscientific Exploration: A Comprehensive Review

The paper analyzed the psychoacoustic challenges of defining timbre, exploring concepts like Timbre Space and the potential of Machine Learning to model neural activity associated with timbre perception. auditory perception.
https://arxiv.org/html/2405.13661v1

An Experiment on DFT - Validation and Application of DFT in Python

A self-initiated study completed prior to formal graduate coursework, focusing on building a solid technical foundation for MIR. This work presents a basic experiment exploring Discrete Fourier Transform (DFT) using Python's NumPy library. It begins by establishing the foundation of the DFT equation and then leverage the NumPy fft package to perform real-world spectrum analysis. While this experiment focuses on fundamental techniques, attempts at pitch detection and filtering offer a glimpse into the vast potential of signal processing.
https://www.sylviastudio.cn/an-experiment-on-dft/

Projects

1. MIDI Player/Analyzer for RAS

A professional-grade MIDI player and analysis suite designed for Rhythmic Auditory Stimulation (RAS) therapy and research. It features two major innovations: a 4-dimensional rhythm analysis framework and a comprehensive batch processing system with multi-method meter classification.

Highlight Features

• Rhythm Analysis: Quantifies rhythm across Beat Density, Rhythm Predictability, Beat Salience, and Rhythmic Uniformity, generating normalized scores for comparative analysis and visualization.
• Batch Processing System: Batch analysis featuring dual-layer meter classification: Tempogram/Autocorrelation-based analysis on MIDI for fast filtering, complemented by RNN-based downbeat tracking on synthesized audio for enhanced precision.
• High-precision Timing System: Microsecond-level accurate metronome with synchronized visual and audio feedback for RAS therapy applications.
• Timing Correction: Estimate the first downbeat and corrects MIDI event timing for synchronization.

Github: https://github.com/tellmeayu/MIDI_Player-Analyzer_for_RAS.git

2. Tranquil Heart (iOS app)

An interactive music rhythm app designed for focus, relaxation and affective modulation. Moving beyond passive listening, it explores Haptic Rhythm -- the concept of using subtle vibration feedback alongside music to cultivate presence and self-awareness.

Interaction Design: Built on iOS (Swift) utilizing simple touch/gesture controls for dynamic interaction. Features three distinct mini-games that integrate Haptics feedback to actively guide user engagement.

Github: https://github.com/tellmeayu/Tranquil_Heart.git

3. Custom DSP-Enabled Sampler Prototype

An early, self-initiated prototype developed in Max/MSP focused on designing a highly personalized, lightweight software sampler.
Technical Highlights: Features a state-variable filter (SVF), custom LFO, and a dedicated signal chain for simulating wave-table synthesis; Implemented sophisticated DSP control logic, including feedback gain compensation for the delay module, and an automated system for clearing residual audio signals.
Engineering Insights: Established a compensation logic to dynamically stablize delay feedback gain, using RMS monitoring. Designed an automated system to clear delay line when the residual signal level dropped below a noise threshold. This is typically to eliminate long, faint noise tails in the audio system. Diagnosed a significant system latency issue when switching MIDI port. Strategically inserted a compensated time delay to ensure a reliable workflow.

Github: https://github.com/tellmeayu/Custom-DSP-Enabled-Sampler-Prototype.git

4. MIDI Feature Visualization & MIDI Utility Script (JavaScript) for Enhanced Protocol Control

Developed a proof-of-concept animation patch in Max/MSP to explore real-time visualization of abstract MIDI data (pitch, velocity, note duration, and track identity).
Demo video: https://youtu.be/9X_WIT-fNKw?si=doXFP0Y7FVYK2F-d
Github: https://github.com/tellmeayu/MIDI-visulisation-Max-MSP-for-Trout-Quintet.git
A further solution-oriented development: A specialized JS script (embedded within a Max/MSP patch) specifically to address the MIDI handling limitations identified during the visualization.
This part of the project was later recognized and featured on the Cycling'74 User Projects website - https://cycling74.com/projects/midi-parsing-using-javascript-in-max .
Github: https://github.com/tellmeayu/MIDI_parsing_thru_JS_in_Max

5. Real-time Beat Tracking and Generative 3D Visualization System

A real-time computational demonstration of algorithmic control over visual outputs in Max/MSP.
Key Features: Algorithmic Structure, Real-time Visualization, HCI
Core function: beat-tracking via logical operations on the input audio signal; Using the derived rhythmic data as the primary control parameter for a generative 3D environment (Jitter/OpenGL).
Demo video: https://youtu.be/wpOkb3GI_JE?si=UvO9nlxpWiopKe10
Github: https://github.com/tellmeayu/Real-time-Beat-Tracking-and-Generative-3D-Visualization-System.git