تحليل الصوت والموسيقى في الوقت الحقيقي عبر مكتبة بايثون مجانية

التثبيت باستخدام Aubio

pip install aubio 

تثبيت Aubio من GitHub

go get -u github.com/aubio/aubio.git 

كيف يتم اكتشاف البدايات في ملف صوتي باستخدام مكتبة بايثون؟

import aubio
import numpy as np

# Configuration
filename = "audio_sample.wav"
win_s = 512        # FFT window size
hop_s = 256        # Hop size (frames between analysis windows)

# Open the audio source
src = aubio.source(filename, hop_size=hop_s)
samplerate = src.samplerate

# Create the onset detector (using 'complex' method)
onset = aubio.onset("complex", win_s, hop_s, samplerate)

# Set detection threshold (0.0 to 1.0 — lower = more sensitive)
onset.set_threshold(0.3)

onsets = []

# Process audio frame by frame
while True:
    samples, read = src()                # Read one hop of audio
    if onset(samples):                   # Returns True when onset detected
        print(f"Onset at: {onset.get_last_s():.3f} seconds")
        onsets.append(onset.get_last_s())
    if read < hop_s:                     # End of file
        break

print(f"\nTotal onsets detected: {len(onsets)}")

كيف يتم تنفيذ اكتشاف النغمة في الوقت الحقيقي باستخدام بايثون؟

import aubio
import numpy as np

samplerate = 44100
win_s = 1024
hop_s = 512

pitch_o = aubio.pitch("yin", win_s, hop_s, samplerate)
pitch_o.set_unit("Hz")
pitch_o.set_silence(-40)

samples = np.random.rand(hop_s).astype(np.float32)
pitch = pitch_o(samples)

print("Detected pitch:", pitch)

كيف يتم قراءة ومعالجة وكتابة ملفات الصوت باستخدام بايثون؟

import aubio
import numpy as np

input_file = "input.wav"
output_file = "output.wav"
target_samplerate = 22050    # Downsample to 22.05 kHz
hop_s = 512

# Open source — resamples automatically to target samplerate
src = aubio.source(input_file, samplerate=target_samplerate, hop_size=hop_s)
actual_sr = src.samplerate
num_channels = src.channels

# Open sink for writing
sink = aubio.sink(output_file, actual_sr, channels=1)

while True:
    samples, read = src()

    # Write mono (sink expects 1D array for mono output)
    if num_channels > 1:
        mono = samples.reshape(num_channels, -1).mean(axis=0)
    else:
        mono = samples

    sink(mono[:read], read)

    if read < hop_s:
        break

del sink   # Flushes and closes file
print(f"Written {output_file} at {actual_sr} Hz")