18-19 HKU CS Final Year Project
Lyrics Recognition for Cantonese Songs
Know More

Project Information

Lyrics Recognition for Cantonese Songs

A python library to identify the lyrics of Cantonese Songs

The objective of the project is to develop a library for identify the Cantonese song lyrics. When the user input an audio file of Cantonese song, the library would process and response text-lyrics corresponding to the song.

The library would be support by a speech recognition model which consists of layers of models: acoustic model, pronounciation model and language model sequentially. These models would be trained by different datasets, the merged model would also be context-dependent.


Au Wing Hing, Terry


So Chi On, Kelvin


Yip Chi Lap, Beta

Project Advisor

Project Timeline

  • Sep, 2018


    • Model Researches
    • Project plan

  • Oct-Nov, 2018

    Requirement Analysis
    Data Preparation

    • Drafted Data Preprocessing Model
    • Raw Training Data

  • Dec 2018

    Model Design

    • Drafted Acoustic Model
    • Adjusted Data Preprocessing Model
    • Processed Data

  • Jan 2019


    • Detailed interim report

  • Feb-Mar 2019


    • Trained Pronunciation Model
    • Trained Language Model
    • Trained acoustic Model
    • Drafted Merged Model

  • Apr 2019

    Testing, Evaluation and Adjustment

    • Adjusted Merged Model

  • Final


The documents will be released based on the project timeline

Project Plan

Click to download

Interim Report

Final Report