从 mp3 到特征：聚类音乐

Question

有没有人有兴趣提供一组 MP3 音乐？我使用了 librosa、ffmpeg 和我编写的 Python 脚本，以便将单个 MP3 文件映射到 114 个数字特征（如速度、mfcc 等）的特征序列。如果对 100.000 个 MP3 重复该过程，则可能会产生一组音乐。问题是如何让人们将他们的 MP3 转换为功能？另一个有趣的问题是，这些功能能否还原为 MP3？这是脚本：

#!/bin/bash

ffmpeg -y -i "" -acodec pcm_u8 -ar 22050 "".wav

python prova.py "".wav

和 Python Wav 功能代码，prova.py:

# Beat tracking example
from __future__ import print_function
import librosa
import numpy as np
import sys


# ffmpeg -i song.mp3 -acodec pcm_u8 -ar 22050 song.wav

# 1. Get the file path to the included audio example
filename = sys.argv[1]
print(filename)
name, extension1, extension2 = filename.split(".")

print("Song name = " + name)

# 2. Load the audio as a waveform `y`
#    Store the sampling rate as `sr`
y, sr = librosa.load(filename)

onset_env = librosa.onset.onset_strength(y, sr=sr)
tempo = librosa.beat.tempo(onset_envelope=onset_env, sr=sr)

# Set the hop length; at 22050 Hz, 512 samples ~= 23ms
hop_length = 512

# Separate harmonics and percussives into two waveforms
y_harmonic, y_percussive = librosa.effects.hpss(y)


#print("y_armonic= " + str(len(y_harmonic)))
#print("y_percussive= " + str(len(y_percussive)))

# Beat track on the percussive signal
tempo, beat_frames = librosa.beat.beat_track(y=y_percussive,
                                             sr=sr)

# Compute MFCC features from the raw signal
mfcc = librosa.feature.mfcc(y=y, sr=sr, hop_length=hop_length, n_mfcc=13)

# And the first-order differences (delta features)
mfcc_delta = librosa.feature.delta(mfcc)

# Stack and synchronize between beat events
# This time, we'll use the mean value (default) instead of median
beat_mfcc_delta = librosa.util.sync(np.vstack([mfcc, mfcc_delta]),
                                    beat_frames)

# Compute chroma features from the harmonic signal
chromagram = librosa.feature.chroma_cqt(y=y_harmonic,
                                        sr=sr)

# Aggregate chroma features between beat events
# We'll use the median value of each feature between beat frames
beat_chroma = librosa.util.sync(chromagram,
                                beat_frames,
                                aggregate=np.median)

# Finally, stack all beat-synchronous features together
beat_features = np.vstack([beat_chroma, beat_mfcc_delta])

print(str(mfcc.shape) + " " + str(mfcc_delta.shape) + " " + str(beat_mfcc_delta.shape) + " " + str(chromagram.shape) + " " + str(beat_chroma.shape) + str(beat_features.shape))

f = open(name + ".txt","w")

f.write(str(tempo)+ ',')

for i in range(0, len(mfcc)):
    f.write(str(np.mean(mfcc[i]))+ ',')

for i in range(0, len(mfcc_delta)):
    f.write(str(np.mean(mfcc_delta[i]))+ ',')

for i in range(0, len(beat_mfcc_delta)):
    f.write(str(np.mean(beat_mfcc_delta[i]))+ ',')

for i in range(0, len(chromagram)):
    f.write(str(np.mean(chromagram[i]))+ ',')

for i in range(0, len(beat_chroma)):
    f.write(str(np.mean(beat_chroma[i]))+ ',')

for i in range(0, len(beat_features)):
    f.write(str(np.mean(beat_features[i]))+ ',')

f.close()

这样，如果将特征集与其流派相结合，给定一首新歌曲，就可以自动判断它属于哪种流派。

Answer 1

要么您提供人们可以在他们的 MP3 上运行的代码（例如命令行工具或桌面应用程序），要么人们为您提供他们的 MP3 供您运行上的代码（例如 Web 应用程序或 HTTP API）。

从 mp3 到特征：聚类音乐

From mp3 to features: clustering music

python

mp3

librosa