| Item | Summary |
|---|---|
| Task | Tomatis-style audio processing / Tomatis 风格音频处理 |
| Core ideas | Dynamic gating, tilt filtering, adaptive control |
| Technical stack | Python, NumPy, SciPy, SoundFile |
| Project value | DSP experimentation, calibration, practical audio engineering |
| Adaptive Spectrum | EQ Curve |
|---|---|
 |
 |
A Python-based audio processing system inspired by Tomatis-style dynamic gated filtering, designed around signal-processing analysis, calibration, and practical experimentation.
一个基于 Python 的 Tomatis 风格音频处理系统，核心是动态门控滤波、频域增益控制和面向真实音频实验的校准流程。
This project started as an attempt to reproduce and experiment with Tomatis-like audio processing behavior in software.
这个项目最初的目标，是把类似 Tomatis 设备的处理逻辑用软件方式复现出来，并进一步支持可调参数、批量实验和音质优化。
It is more than a simple filter implementation:
- dynamic gating between different processing states
- tilt-style frequency shaping
- adaptive parameter tuning
- practical engineering for long audio files and boundary handling
它不只是一个简单滤波器，而是一整套：
- 动态门控状态切换
- 倾斜式频率增益控制
- 自适应参数优化
- 面向长音频和真实边界问题的工程处理
The processor switches between different filter states based on input signal level.
处理器会根据输入音频电平在不同滤波状态之间切换，以模拟更动态的听觉处理行为。
Frequency-dependent gain shaping is applied to emphasize or suppress different frequency regions.
在频域中对不同频率段做倾斜式增益控制，从而实现低频/高频的动态强调与抑制。
In addition to a more device-like mode, the project includes an adaptive variant for smoother transitions and threshold optimization.
除了更接近设备复刻的模式外，这个项目还包含自适应版本，用于改善切换平滑性并自动优化阈值。
The implementation includes practical handling for:
- STFT + OLA processing
- boundary padding
- clipping avoidance
- transition smoothing
- long-file processing workflows
实现中包含大量真实工程细节，例如：
- STFT + OLA
- 边界 padding
- 防止削波
- 平滑过渡
- 长音频处理流程
- Python 3.11+
- NumPy
- SciPy
- SoundFile
- Librosa
- Pandas
- Matplotlib
python src/process_tomatis.py -i input.flac -o output.flac --gate_ui 50python src/process_tomatis_adaptive.py -i input.flac -o output.flacThis project is a good example of work that sits between:
- DSP experimentation
- reverse-engineering style thinking
- practical Python implementation
- iterative calibration and debugging
这个项目比较有辨识度，因为它处在下面几个方向的交叉点：
- 数字信号处理实验
- 类逆向工程式的分析思路
- Python 工程实现
- 反复校准和调参的实践过程
It shows how I approach systems that are not fully specified upfront: I test behavior, compare outputs, refine assumptions, and gradually turn that into a working implementation.
它体现了我处理“规范不完整系统”的方式：先观察行为、做对比、修正假设，再逐步把它沉淀成可运行实现。
src/ # core processing logic
scripts/ # utility and analysis scripts
docs/ # technical notes and work logs
analysis/ # experiment outputs and supporting analysis
audio/ # local audio assets (not all suitable for publication)
docs/Tomatis处理器使用指南.mddocs/Tomatis技术说明.mddocs/TOMATIS_WORK_LOG.md
These documents capture the algorithm intuition, parameter choices, and implementation evolution.
这些文档记录了算法思路、参数设计和项目演进过程。
This repository is best viewed as a project-and-experiment repository rather than a polished library package.
本仓库更适合作为一个“项目 + 实验记录”仓库来理解，而不是已经彻底产品化的库。
If I continue refining it, the next steps would be:
- cleanly separating reusable processing code from one-off calibration scripts
- adding smaller demo audio samples
- packaging a more consistent CLI interface
- adding tests for core DSP utilities
如果后续继续整理，我会优先：
- 把可复用处理代码和一次性校准脚本拆开
- 增加小型 demo 音频样本
- 统一 CLI 入口
- 为核心 DSP 工具补充测试
This project started as an attempt to reproduce and experiment with Tomatis-like audio processing behavior in software.
这个项目最初的目标，是把类似 Tomatis 设备的处理逻辑用软件方式复现出来，并进一步支持可调参数、批量实验和音质优化。
It is more than a simple filter implementation:
- dynamic gating between different processing states
- tilt-style frequency shaping
- adaptive parameter tuning
- practical engineering for long audio files and boundary handling
它不只是一个简单滤波器，而是一整套：
- 动态门控状态切换
- 倾斜式频率增益控制
- 自适应参数优化
- 面向长音频和真实边界问题的工程处理
The processor switches between different filter states based on input signal level.
处理器会根据输入音频电平在不同滤波状态之间切换，以模拟更动态的听觉处理行为。
Frequency-dependent gain shaping is applied to emphasize or suppress different frequency regions.
在频域中对不同频率段做倾斜式增益控制，从而实现低频/高频的动态强调与抑制。
In addition to a more device-like mode, the project includes an adaptive variant for smoother transitions and threshold optimization.
除了更接近设备复刻的模式外，这个项目还包含自适应版本，用于改善切换平滑性并自动优化阈值。
The implementation includes practical handling for:
- STFT + OLA processing
- boundary padding
- clipping avoidance
- transition smoothing
- long-file processing workflows
实现中包含大量真实工程细节，例如：
- STFT + OLA
- 边界 padding
- 防止削波
- 平滑过渡
- 长音频处理流程
- Python 3.11+
- NumPy
- SciPy
- SoundFile
- Librosa
- Pandas
- Matplotlib
python src/process_tomatis.py -i input.flac -o output.flac --gate_ui 50python src/process_tomatis_adaptive.py -i input.flac -o output.flacThis project is a good example of work that sits between:
- DSP experimentation
- reverse-engineering style thinking
- practical Python implementation
- iterative calibration and debugging
这个项目比较有辨识度，因为它处在下面几个方向的交叉点：
- 数字信号处理实验
- 类逆向工程式的分析思路
- Python 工程实现
- 反复校准和调参的实践过程
It shows how I approach systems that are not fully specified upfront: I test behavior, compare outputs, refine assumptions, and gradually turn that into a working implementation.
它体现了我处理“规范不完整系统”的方式：先观察行为、做对比、修正假设，再逐步把它沉淀成可运行实现。
src/ # core processing logic
scripts/ # utility and analysis scripts
docs/ # technical notes and work logs
analysis/ # experiment outputs and supporting analysis
audio/ # local audio assets (not all suitable for publication)
docs/Tomatis处理器使用指南.mddocs/Tomatis技术说明.mddocs/TOMATIS_WORK_LOG.md
These documents capture the algorithm intuition, parameter choices, and implementation evolution.
这些文档记录了算法思路、参数设计和项目演进过程。
This repository is best viewed as a project-and-experiment repository rather than a polished library package.
本仓库更适合作为一个“项目 + 实验记录”仓库来理解，而不是已经彻底产品化的库。
If I continue refining it, the next steps would be:
- cleanly separating reusable processing code from one-off calibration scripts
- adding smaller demo audio samples
- packaging a more consistent CLI interface
- adding tests for core DSP utilities
如果后续继续整理，我会优先：
- 把可复用处理代码和一次性校准脚本拆开
- 增加小型 demo 音频样本
- 统一 CLI 入口
- 为核心 DSP 工具补充测试