如何绘制两个WAV文件的频谱子图？颜色条设置与图幅调整

一、先修正原代码的核心错误 #

你当前代码里有个关键问题：两个子图用的是同一个S_dB（仅对应最后加载的音频），导致两张图显示完全一样。需要分别处理两个音频的梅尔频谱：

# 处理无说话音频
y_no_speak, sr = librosa.load(FILE_NO_SPEAKING)
mel_spec_no_speak = librosa.feature.melspectrogram(y=y_no_speak, sr=sr)
S_dB_no_speak = librosa.power_to_db(mel_spec_no_speak, ref=np.max)

# 处理有说话音频
y_speak, sr = librosa.load(FILE_SPEAKING)
mel_spec_speak = librosa.feature.melspectrogram(y=y_speak, sr=sr)
S_dB_speak = librosa.power_to_db(mel_spec_speak, ref=np.max)

二、放大绘图整体尺寸 #

创建子图时，通过figsize参数直接设置画布宽高（数值越大，图的尺寸越大）：

# 示例：宽16，高6的画布，1行2列子图
fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(16, 6))

三、配置颜色条 #

有两种实用方案可选：

方案1：每个子图单独配颜色条 #

分别为两个子图绑定颜色条，需要传入specshow返回的图像对象：

# 绘制无说话频谱图并加颜色条
img_no_speak = librosa.display.specshow(S_dB_no_speak, 
                                        x_axis='time',
                                        y_axis='mel', 
                                        sr=sr,
                                        fmax=8000, 
                                        ax=ax1)
fig.colorbar(img_no_speak, ax=ax1, format='%+2.0f dB')

# 绘制有说话频谱图并加颜色条
img_speak = librosa.display.specshow(S_dB_speak, 
                                     x_axis='time',
                                     y_axis='mel', 
                                     sr=sr,
                                     fmax=8000, 
                                     ax=ax2)
fig.colorbar(img_speak, ax=ax2, format='%+2.0f dB')

方案2：共享一个颜色条（更适合对比） #

如果两个音频的分贝范围一致，共享颜色条能让对比更直观，把颜色条放在两个子图右侧：

# 绘制两个频谱图
img_no_speak = librosa.display.specshow(S_dB_no_speak, 
                                        x_axis='time',
                                        y_axis='mel', 
                                        sr=sr,
                                        fmax=8000, 
                                        ax=ax1)
img_speak = librosa.display.specshow(S_dB_speak, 
                                     x_axis='time',
                                     y_axis='mel', 
                                     sr=sr,
                                     fmax=8000, 
                                     ax=ax2)

# 绑定到两个子图，共享颜色条
fig.colorbar(img_speak, ax=[ax1, ax2], format='%+2.0f dB')

完整修正后的代码 #

import librosa
import librosa.display
import numpy as np
import matplotlib.pyplot as plt

# 替换为你的实际文件路径
FILE_NO_SPEAKING = "no_speaking.wav"
FILE_SPEAKING = "speaking.wav"

# 加载并处理音频
y_no_speak, sr = librosa.load(FILE_NO_SPEAKING)
mel_spec_no_speak = librosa.feature.melspectrogram(y=y_no_speak, sr=sr)
S_dB_no_speak = librosa.power_to_db(mel_spec_no_speak, ref=np.max)

y_speak, sr = librosa.load(FILE_SPEAKING)
mel_spec_speak = librosa.feature.melspectrogram(y=y_speak, sr=sr)
S_dB_speak = librosa.power_to_db(mel_spec_speak, ref=np.max)

# 创建指定尺寸的画布
fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(16, 6))

# 绘制第一个子图
img_no_speak = librosa.display.specshow(S_dB_no_speak, 
                                        x_axis='time',
                                        y_axis='mel', 
                                        sr=sr,
                                        fmax=8000, 
                                        ax=ax1)
fig.colorbar(img_no_speak, ax=ax1, format='%+2.0f dB')
ax1.set(title='No Speak')

# 绘制第二个子图
img_speak = librosa.display.specshow(S_dB_speak, 
                                     x_axis='time',
                                     y_axis='mel', 
                                     sr=sr,
                                     fmax=8000, 
                                     ax=ax2)
fig.colorbar(img_speak, ax=ax2, format='%+2.0f dB')
ax2.set(title='Speak')

# 调整布局避免标签重叠
plt.tight_layout()
plt.show()

内容的提问来源于stack exchange，提问作者user3668129