Merge pull request #34 from wtq2255/master

update v0.1.8

LICENSE.md

@@ -1,6 +1,6 @@
 ## MIT License
 
-Copyright (c) 2021--2023, `audioFlux` development team.
+Copyright (c) 2021--2024, `audioFlux` development team.
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

README.md

@@ -2,17 +2,6 @@
 
 # audioFlux
 
-<!--<p align="center">
-<img src='./image/logo.png'  width="380"  style="max-width: 100%;" > 
-</p>-->
-
-
-<!-- 
-[![Pypi Downloads](https://img.shields.io/pypi/dm/aubio.svg?label=Pypi%20downloads)](https://pypi.org/project/aubio/)
-[![Conda Downloads](https://img.shields.io/conda/dn/conda-forge/aubio.svg?label=Conda%20downloads)](https://anaconda.org/conda-forge/aubio)
-[![Documentation](https://readthedocs.org/projects/aubio/badge/?version=latest)](http://aubio.readthedocs.io/en/latest/?badge=latest "Latest documentation") -->
-<!--![GitHub code size in bytes](https://img.shields.io/github/languages/code-size/libAudioFlux/audioFlux)-->
-
 <!-- shields.io -->
 ![GitHub Workflow Status (with branch)](https://img.shields.io/github/actions/workflow/status/libAudioFlux/audioFlux/build.yml?branch=master)
 ![example branch parameter](https://github.com/libAudioFlux/audioFlux/actions/workflows/build.yml/badge.svg?branch=master)
@@ -39,6 +28,13 @@ field such as Classification, Separation, Music Information Retrieval(MIR) and A
 - High performance, core part C implementation, FFT hardware acceleration based on different platforms, convenient for large-scale data feature extraction.
 - It supports the mobile end and meets the real-time calculation of audio stream at the mobile end. -->
 
+
+##### New Features
+* v0.1.8
+  * Add a variety of Pitch algorithms: `YIN`, `CEP`, `PEF`, `NCF`, `HPS`, `LHS`, `STFT` and `FFP`.
+  * Add `PitchShift` and `TimeStretch` algorithms.
+
+
 ### Table of Contents
 
 - [Overview](#overview)

conda/meta.yaml

@@ -1,4 +1,4 @@
-{% set version = "0.1.7" %}
+{% set version = "0.1.8" %}
 
 package:
   name: audioflux

docs/_static/versions.json

@@ -1,9 +1,14 @@
 [
     {
-        "name": "0.1.7 (stable)",
-        "version": "0.1.7",
+        "name": "0.1.8 (stable)",
+        "version": "0.1.8",
         "url": "https://audioflux.top/stable/"
     },
+    {
+        "name": "0.1.7",
+        "version": "0.1.7",
+        "url": "https://audioflux.top/v0.1.7/"
+    },
     {
         "name": "0.1.6",
         "version": "0.1.6",

docs/changelog.rst

@@ -1,9 +1,20 @@
 ChangeLog
 =========
+v0.1.8
+------
+* New features:
+    * Add `audioflux.STFT` algorithm.
+    * Add `audioflux.PitchFFP` algorithm.
+    * Add `audioflux.PitchShift` and `audioflux.TimeStretch` algorithms.
+    * Add method `temporal` to `audioflux.Temporal`.
+    * Optimize performance.
+* Fix bug:
+    * Fix `nmf` bug.
+
 v0.1.7
 ------
 * New features:
-    * Add Cepstrogram algorithm.
+    * Add `audioflux.Cepstrogram` algorithm.
     * Add PitchCEP/PitchHPS/PitchLHS/PitchNCF/PitchPEF/PitchSTFT/PitchYIN algorithm.
 * Modified API:
     * `audioflux.display.fill_wave` add `times` params.

docs/installation.rst

@@ -79,10 +79,10 @@ Linux build
     # For CentOS:
     $ sudo yum install -y cmake clang
 
-2. Installing **MKL** lib dependencies on Linux
+2. Installing **MKL** lib dependencies on Linux (Optional, Recommended installation)
 
    You can
-   use `this installation document <https://www.intel.cn/content/www/cn/zh/developer/tools/oneapi/onemkl-download.html?operatingsystem=linux>`_
+   use `this installation document <https://www.intel.com/content/www/us/en/developer/tools/oneapi/onemkl-download.html?operatingsystem=linux>`_
    to install MKL
 
    After installation, you need to set the environment variable `MKL_INCLUDE_PATH` and `MKL_LIB_PATH` for MKL.
@@ -128,8 +128,9 @@ macOS build
 Windows build
 ^^^^^^^^^^^^^
 
-Building from source is currently not supported. Only supports pip installation. If you need to build from source, The
-easiest way to build audioFlux is by cross-compilation on Linux/macOS using MinGW::
+Building from source code is currently not supported on Windows; it is recommended to use pip installation.
+
+If you need to build from source, The easiest way to build audioFlux is by cross-compilation on Linux/macOS using MinGW.::
 
  $ python setup.py build_py_win
  $ python setup.py install

docs/installing.md

@@ -84,10 +84,10 @@ Build and compile successfully, the project build compilation results are in the
     $ sudo yum install -y cmake clang
     ```
 
-2. Installing **MKL** lib dependencies on Linux
+2. Installing **MKL** lib dependencies on Linux (Optional, Recommended installation)
 
    You can
-   use [this installation document](https://www.intel.cn/content/www/cn/zh/developer/tools/oneapi/onemkl-download.html?operatingsystem=linux)
+   use [this installation document](https://www.intel.com/content/www/us/en/developer/tools/oneapi/onemkl-download.html?operatingsystem=linux)
    to install MKL
 
    After installation, you need to set the environment variable `MKL_INCLUDE_PATH` and `MKL_LIB_PATH` for MKL.
@@ -151,8 +151,9 @@ Build and compile successfully, the project build compilation results are in the
 
 #### Windows build
 
-Building from source is currently not supported. Only supports pip installation. If you need to build from source, The
-easiest way to build audioFlux is by cross-compilation on Linux/macOS using MinGW.
+Building from source code is currently not supported on Windows; it is recommended to use pip installation.
+
+If you need to build from source, The easiest way to build audioFlux is by cross-compilation on Linux/macOS using MinGW.
 
  ```shell
  $ python setup.py build_py_win

docs/mir/harmonics.rst

@@ -0,0 +1,8 @@
+Harmonics
+=========
+
+.. autoclass:: audioflux.Harmonic
+   :members:
+
+.. autoclass:: audioflux.HarmonicRatio
+   :members:

docs/mir/mir.rst

@@ -9,6 +9,10 @@ The mir module contains the following algorithms:
     pitch
     onsetBase
     hpss
+    harmonics
+    pitchShift
+    timeStretch
+
 
 +-----------------+---------------------------------+
 | MIR             | algorithm                       |

docs/mir/pitch.rst

@@ -1,6 +1,12 @@
 Pitch
 =====
 
+.. autoclass:: audioflux.PitchYIN
+   :members:
+
+.. autoclass:: audioflux.PitchPEF
+   :members:
+
 .. autoclass:: audioflux.PitchCEP
    :members:
 
@@ -13,11 +19,8 @@ Pitch
 .. autoclass:: audioflux.PitchNCF
    :members:
 
-.. autoclass:: audioflux.PitchPEF
-   :members:
-
 .. autoclass:: audioflux.PitchSTFT
    :members:
 
-.. autoclass:: audioflux.PitchYIN
+.. autoclass:: audioflux.PitchFFP
    :members:

docs/mir/pitchShift.rst

@@ -0,0 +1,5 @@
+PitchShift
+==========
+
+.. autoclass:: audioflux.PitchShift
+   :members:

docs/mir/timeStretch.rst

@@ -0,0 +1,5 @@
+TimeStretch
+===========
+
+.. autoclass:: audioflux.TimeStretch
+   :members:

docs/transforms/stft.rst

@@ -0,0 +1,5 @@
+STFT - Short Time Fourier Transform
+===================================
+
+.. autoclass:: audioflux.STFT
+   :members:

docs/transforms/transforms.rst

@@ -51,6 +51,7 @@ The following transform are not supports multiple frequency scale types, only us
     dwt
     wpt
     swt
+    stft
 
 
 Synchronized Squeezing

python/audioflux/__init__.py

@@ -10,6 +10,7 @@
 from .pwt import *
 from .reassign import *
 from .st import *
+from .stft import *
 from .swt import *
 from .synsq import *
 from .temporal import *
@@ -24,6 +25,9 @@
 from .feature import *
 from .core import *
 from .spectrogram import MelSpectrogram, BarkSpectrogram, ErbSpectrogram
+from . import mir
+from . import feature
+from . import dsp
 from . import utils
 from . import type
 from . import display

python/audioflux/__version__.py

@@ -1,3 +1,3 @@
 __title__ = 'audioflux'
 __description__ = 'A library for audio and music analysis, feature extraction.'
-__version__ = '0.1.7'
+__version__ = '0.1.8'

python/audioflux/bft.py

@@ -231,7 +231,7 @@ def cal_time_length(self, data_length):
         Calculate the length of a frame from audio data.
 
         - ``fft_length = 2 ** radix2_exp``
-        - ``(data_length - fft_length) / slide_length + 1``
+        - ``(data_length - fft_length) // slide_length + 1``
 
         Parameters
         ----------
@@ -444,25 +444,25 @@ def _get_temporal_data(self, data_length):
 
         fn = self._lib['bftObj_getTemporalData']
         fn.argtypes = [POINTER(OpaqueBFT),
-                       POINTER(POINTER(c_float)),
-                       POINTER(POINTER(c_float)),
-                       POINTER(POINTER(c_float)),
+                       POINTER(c_void_p),
+                       POINTER(c_void_p),
+                       POINTER(c_void_p),
                        ]
         time_length = self.cal_time_length(data_length)
 
-        pp_energy_arr = pointer(pointer(c_float()))
-        pp_rms_arr = pointer(pointer(c_float()))
-        pp_zcr_arr = pointer(pointer(c_float()))
+        pp_energy_arr = pointer(c_void_p())
+        pp_rms_arr = pointer(c_void_p())
+        pp_zcr_arr = pointer(c_void_p())
 
         fn(self._obj,
            pp_energy_arr,
            pp_rms_arr,
            pp_zcr_arr
            )
 
-        energy_arr = np.array([pp_energy_arr.contents[x] for x in range(time_length)], dtype=np.float32)
-        rms_arr = np.array([pp_rms_arr.contents[x] for x in range(time_length)], dtype=np.float32)
-        zcr_arr = np.array([pp_zcr_arr.contents[x] for x in range(time_length)], dtype=np.float32)
+        energy_arr = np.frombuffer((c_float * time_length).from_address(pp_energy_arr[0]), np.float32).copy()
+        rms_arr = np.frombuffer((c_float * time_length).from_address(pp_rms_arr[0]), np.float32).copy()
+        zcr_arr = np.frombuffer((c_float * time_length).from_address(pp_zcr_arr[0]), np.float32).copy()
 
         return energy_arr, rms_arr, zcr_arr
 

python/audioflux/classic/__init__.py

@@ -0,0 +1 @@
+from .nmf import *

python/audioflux/classic/nmf.py

@@ -0,0 +1,37 @@
+from ctypes import c_int, POINTER, c_float, pointer
+
+import numpy as np
+
+from audioflux.fftlib import get_fft_lib
+
+__all__ = [
+    'nmf',
+]
+
+
+def nmf(X, k, max_iter=300, tp=0, thresh=1e-3, norm=0):
+    X = np.asarray(X, dtype=np.float32, order='C')
+    if X.ndim != 2:
+        raise ValueError(f"X[ndim={X.ndim}] must be a 2D array")
+
+    fn = get_fft_lib()['nmf']
+    fn.argtypes = [
+        np.ctypeslib.ndpointer(dtype=np.float32, ndim=2, flags='C_CONTIGUOUS'),
+        c_int, c_int, c_int,
+        np.ctypeslib.ndpointer(dtype=np.float32, ndim=2, flags='C_CONTIGUOUS'),
+        np.ctypeslib.ndpointer(dtype=np.float32, ndim=2, flags='C_CONTIGUOUS'),
+        POINTER(c_int), POINTER(c_int),
+        POINTER(c_float), POINTER(c_int)
+    ]
+
+    n_len, m_len = X.shape
+
+    h_arr = np.arange(1, k * m_len + 1, dtype=np.float32).reshape((k, m_len))
+    w_arr = np.arange(1, n_len * k + 1, dtype=np.float32).reshape((n_len, k))
+
+    fn(X, c_int(n_len), c_int(m_len), c_int(k),
+       w_arr, h_arr,
+       pointer(c_int(max_iter)), pointer(c_int(tp)),
+       pointer(c_float(thresh)), pointer(c_int(norm)))
+
+    return h_arr, w_arr

python/audioflux/display/display.py

@@ -82,24 +82,20 @@ class TimeFormatter(Formatter):
     def __init__(self, lag=False, unit=None):
 
         if unit not in ["s", "ms", None]:
-            raise ValueError("Unknown time unit: {}".format(unit))
+            raise ValueError("Time unit is not support: {}".format(unit))
 
         self.unit = unit
         self.lag = lag
 
     def __call__(self, x, pos=None):
-        """Return the time format as pos"""
 
         _, dmax = self.axis.get_data_interval()
         vmin, vmax = self.axis.get_view_interval()
 
-        # In lag-time axes, anything greater than dmax / 2 is negative time
         if self.lag and x >= dmax * 0.5:
-            # In lag mode, don't tick past the limits of the data
             if x > dmax:
                 return ""
             value = np.abs(x - dmax)
-            # Do we need to tweak vmin/vmax here?
             sign = "-"
         else:
             value = x
@@ -111,7 +107,6 @@ def __call__(self, x, pos=None):
             s = "{:.3g}".format(value * 1000)
         else:
             if vmax - vmin > 3600:
-                # Hours viz
                 s = "{:d}:{:02d}:{:02d}".format(
                     int(value / 3600.0),
                     int(np.mod(value / 60.0, 60)),
@@ -121,10 +116,8 @@ def __call__(self, x, pos=None):
                 # Minutes viz
                 s = "{:d}:{:02d}".format(int(value / 60.0), int(np.mod(value, 60)))
             elif vmax - vmin >= 1:
-                # Seconds viz
                 s = "{:.2g}".format(value)
             else:
-                # Milliseconds viz
                 s = "{:.3f}".format(value)
 
         return "{:s}{:s}".format(sign, s)
@@ -274,7 +267,7 @@ def fill_plot(x, y, axes=None, label='', is_legend=True, *,
     axes.set_ylim(*y_lims)
 
     axes.plot(x, y, label=label)
-    if is_legend:
+    if is_legend and label:
         axes.legend()
     return axes
 

python/audioflux/dsp/__init__.py

@@ -1,2 +1,3 @@
 from .czt import *
 from .resample import *
+from .xcorr import *