Real-time Pitch Adjustment on BF533 DSP

This repository shows my course project in Systems Design based on Digital Signal Processors, 2022-2023, Department of Electronic Engineering by Professor Weibei Dou.

The main idea to adjustment pitch of the audio is to:

interpolate and resample the audio
split and overlap-add frames using a gap equals $\frac{H_s}{H_a}$

Python

Python codes are used to verify the feasibility of the algorithm, which is not real-time.

Setup Development Environment

$ conda create -n audioexp python=3.10 pip
$ conda activate audioexp
$ pip install numpy scipy

Resample and Overlap-add (OLA)

Create a folder output in Python/.
Modify parameters such as wav_name and alpha in resample-stretch.py.
Run python resample-stretch.py
Play audios generated in Python/output/.

C

C codes are based on demo project Talk_Through_533_I2S _FIR_2MAC.rar provided by Systems Design based on Digital Signal Processors. It aujusts pitch of the input audio in real-time.

Create the Project

Open IDDE CrossCore®Embedded Studio 2.10.1 and create a new project.
Choose BF533 with Silicon Revision any selected. Do not add Startup Code/LDF.
In Project > properties > C/C++ Build > Settings:
- CrossCore Blackfin Linker > Processor > Tool Settings > Memory usage, select Internal SRAM and external SDRAM. In
- CrossCore Blackfin Linker > General > Tool Settings, select Individually map functions and data items.
Rename C folder in this repo to src to replace the scr folder generate by IDDE.
Build and debug the project.

Contribute

Issues and PRs are welcome!

References

A Review of Time-Scale Modification of Music Signals, Jonathan Driedger and Meinard Müller, International Audio Laboratories Erlangen, 91058 Erlangen, Germany
变声导论-变声器原理及实现(核心算法实现篇) Dbinary@知乎
变声导论-变声器原理及实现(基音分类与滤波系统实现) Dbinary@知乎
语音变调讲解PPT 潘兴 2022.10.20
基于数字信号处理器的系统设计课程课件窦维蓓
基于数字信号处理器的系统设计课程设计选题指南窦维蓓戴喜增周斌余吉张波何以凡等
SoundLab

Yang-Xijie / PitchAdjustmentDSP