modusa 0.4.30__py3-none-any.whl → 0.4.31__py3-none-any.whl

modusa/__init__.py

@@ -11,4 +11,7 @@ from modusa.tools import dist_plot, hill_plot, plot, fig
 # Synthsizing related
 from modusa.tools import synth_f0
 
-__version__ = "0.4.30" # This is dynamically used by the documentation, and pyproject.toml; Only need to change it here; rest gets taken care of.
+# Audio features related
+from modusa.tools import stft
+
+__version__ = "0.4.31" # This is dynamically used by the documentation, and pyproject.toml; Only need to change it here; rest gets taken care of.

modusa/tools/__init__.py

@@ -16,4 +16,7 @@ from .plotter import Fig as fig
 from .plotter import dist_plot, hill_plot, plot
 
 # Synthesizing related
-from .synth import synth_f0
+from .synth import synth_f0
+
+# Audio features
+from .audio_stft import stft

modusa/tools/audio_stft.py

@@ -0,0 +1,72 @@
+#!/usr/bin/env python3
+
+#---------------------------------
+# Author: Ankit Anand
+# Date: 23/10/25
+# Email: ankit0.anand0@gmail.com
+#---------------------------------
+
+import numpy as np
+
+def stft(y, sr, winlen=None, hoplen=None, gamma=None):
+	"""
+	Compute spectrogram with just numpy.
+
+	Parameters
+	----------
+	y: ndarray
+		- Audio signal.
+	sr: int
+		- Sampling rate of the audio signal.
+	winlen: int
+		- Window length in samples.
+		- Default: None => set at 0.064 sec
+	hoplen: int
+		- Hop length in samples.
+		- Default: None => set at one-forth of winlen
+	gamma: int | None
+		- Log compression factor.
+		- Add contrast to the plot.
+
+	Returns
+	-------
+	ndarray:
+		- Spectrogram matrix, complex is gamma is None else real
+	ndarray:
+		- Frequency bins in Hz.
+	ndarray:
+		- Timeframes in sec.
+	"""
+	
+	if winlen is None:
+		winlen = 2 ** int(np.log2(0.064 * sr))
+	if hoplen is None:
+		hoplen = int(winlen * 0.25)
+		
+	# Estimating the shape of the S matrix
+	M = int(np.ceil(winlen / 2))
+	N = int(np.ceil((y.size - winlen) / hoplen))
+	
+	# Initialize the S matrix
+	S = np.empty((M, N), dtype=np.complex64) # M X N => freq bin X time frame
+	
+	# We will need a hann window
+	hann = np.hanning(winlen)
+	
+	# Get the frames
+	frames = np.lib.stride_tricks.sliding_window_view(y, window_shape=winlen)[::hoplen] # frame X chunk
+	
+	# Apply window to the frame
+	frames_windowed = frames * hann # frame X chunk
+	
+	# Compute fft for each frame
+	S = np.fft.rfft(frames_windowed, n=winlen, axis=1).T  # transpose to match shape (freq_bins, frame)
+	
+	# Magnitude spectrogram
+	Sf = np.fft.rfftfreq(winlen, d=1/sr) # Frequency bins (Hz)
+	St = np.arange(N) * hoplen / sr # Time bins (sec)
+	
+	if gamma is not None:
+		S = np.log1p(gamma * np.abs(S))
+		
+	return S, Sf, St

{modusa-0.4.30.dist-info → modusa-0.4.31.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: modusa
-Version: 0.4.30
+Version: 0.4.31
 Summary: A modular signal analysis python library.
 Author-Email: Ankit Anand <ankit0.anand0@gmail.com>
 License: MIT

{modusa-0.4.30.dist-info → modusa-0.4.31.dist-info}/RECORD

@@ -1,21 +1,22 @@
 LICENSE.md,sha256=JTaXAjx5awk76VArKCx5dUW8vmLEWsL_ZlR7-umaHbA,1078
 README.md,sha256=Ybe2rcDecfSCiUpec2n7btQvgyi4R9JG0bfdwSWijWk,981
-modusa-0.4.30.dist-info/METADATA,sha256=6uwfk2M5n8Nm403plD97tZhsFlDjsrCWCGe5Q6uViLM,1443
-modusa-0.4.30.dist-info/WHEEL,sha256=9P2ygRxDrTJz3gsagc0Z96ukrxjr-LFBGOgv3AuKlCA,90
-modusa-0.4.30.dist-info/entry_points.txt,sha256=fmKpleVXj6CdaBVL14WoEy6xx7JQCs85jvzwTi3lePM,73
-modusa-0.4.30.dist-info/licenses/LICENSE.md,sha256=JTaXAjx5awk76VArKCx5dUW8vmLEWsL_ZlR7-umaHbA,1078
-modusa/__init__.py,sha256=Q5A97PDmbcNLEwJZKg3W660TcVyJbbJfTvjct8Y_X9s,450
+modusa-0.4.31.dist-info/METADATA,sha256=6xtxTdA9YYiKzG3pMSmeMk8C18gvlWS7ZJ0GuddrESY,1443
+modusa-0.4.31.dist-info/WHEEL,sha256=9P2ygRxDrTJz3gsagc0Z96ukrxjr-LFBGOgv3AuKlCA,90
+modusa-0.4.31.dist-info/entry_points.txt,sha256=fmKpleVXj6CdaBVL14WoEy6xx7JQCs85jvzwTi3lePM,73
+modusa-0.4.31.dist-info/licenses/LICENSE.md,sha256=JTaXAjx5awk76VArKCx5dUW8vmLEWsL_ZlR7-umaHbA,1078
+modusa/__init__.py,sha256=OlPmjLOhldc42ytXjXkv0zMoJd8B4lndGcG2Pt2olLc,506
 modusa/fonts/NotoSansDevanagari-Regular.ttf,sha256=CEqU2J61Sq-5OgVuFUJcNP2Fn2NCh1Fl0wSDezvPwtI,221084
 modusa/images/icon.png,sha256=UgGXlL4xjdIYxqO1en3iQfNN1BFsazxpT7NiPZSDu70,13321
-modusa/tools/__init__.py,sha256=9tIk1j1lMJ8Ka04ppLwEad_RU-s1Ok9CRalfNhai01A,453
+modusa/tools/__init__.py,sha256=08f7kU1ZTuvYuMTYodv_OK_29AeCspPu4FzTVbbnPm4,500
 modusa/tools/ann_loader.py,sha256=m6Qu6jXnQ8LfUhKItoHSaHlGxUyzUJlGEyu4_50qJ8w,3099
 modusa/tools/ann_saver.py,sha256=rB2gP1rD69PkSnBZrzc_83ylb7eAyqyCMjn9Bh16VaM,650
 modusa/tools/audio_converter.py,sha256=415qBoPm2sBIuBSI7m1XBKm0AbmVmPydIPPr-uO8D3c,1778
 modusa/tools/audio_loader.py,sha256=xmeodrJqJV5j9-lrUPk3W9-bdEA4-6RvTAn8pIErUN0,4084
 modusa/tools/audio_player.py,sha256=kyBUnodkOE9Ox-hKHkfPeGAQ1RPTddbZYXO1ezz6-9w,2494
 modusa/tools/audio_recorder.py,sha256=cADeeUpPcDIYBIuuVZvWqulDkv7TavkpL3o3SO9QyKc,2787
+modusa/tools/audio_stft.py,sha256=vy1J-qFUVIkOh5Y1-dyy3zAh26n0IaMtO2D9EMbtITU,1742
 modusa/tools/plotter.py,sha256=QfF8kCi79nS-J5lQyEuH_BbAmTo-PdfjDbnK9yTAmRE,30863
 modusa/tools/synth.py,sha256=tfCjTAucTDwb0yOHHnrNO7EFsuH7tNb1PwvS2xwmhK0,1222
 modusa/tools/youtube_downloader.py,sha256=Ij7fipSlRpsf0pFOwRY-j7Yf2PMp0kpWMhyMQchR0e0,1574
-pyproject.toml,sha256=zBJWAUi45I8BE-MxTX-ZYMBzNpvhiYE8zcKUviuHNFc,1414
-modusa-0.4.30.dist-info/RECORD,,
+pyproject.toml,sha256=fCgN1j7fzWdYCCNut4pUfqpSfGccBEeGc8pjRXyz0RM,1414
+modusa-0.4.31.dist-info/RECORD,,

pyproject.toml

@@ -16,7 +16,7 @@ dependencies = [
 ]
 requires-python = ">=3.11"
 readme = "README.md"
-version = "0.4.30"
+version = "0.4.31"
 
 [project.license]
 text = "MIT"