neverlib 0.2.5__py3-none-any.whl → 0.2.7__py3-none-any.whl

neverlib/data_analyze/visualization.py

@@ -18,37 +18,40 @@ from scipy.signal import spectrogram
 
 class AudioVisualizer:
     """音频可视化器类"""
-    
+
     def __init__(self, sr: int = 22050, figsize: Tuple[int, int] = (12, 8)):
         """
         初始化可视化器
-        
+
         Args:
             sr: 采样率
             figsize: 图形大小
         """
         self.sr = sr
         self.figsize = figsize
-        
+
         # 设置中文字体支持
         plt.rcParams['font.sans-serif'] = ['SimHei', 'DejaVu Sans']
         plt.rcParams['axes.unicode_minus'] = False
-        
+
         # 设置样式
         plt.style.use('default')
         sns.set_palette("husl")
-    
-    def plot_waveform(self, audio: np.ndarray, title: str = "音频波形图", 
-                     show_time: bool = True, ax: Optional[plt.Axes] = None) -> plt.Figure:
+
+    def plot_waveform(self,
+                      audio: np.ndarray,
+                      title: str = "音频波形图",
+                      show_time: bool = True,
+                      ax: Optional[plt.Axes] = None) -> plt.Figure:
         """
         绘制音频波形图
-        
+
         Args:
             audio: 音频信号
             title: 图标题
             show_time: 是否显示时间轴
             ax: matplotlib轴对象
-            
+
         Returns:
             图形对象
         """
@@ -56,7 +59,7 @@ class AudioVisualizer:
             fig, ax = plt.subplots(figsize=self.figsize)
         else:
             fig = ax.figure
-        
+
         if show_time:
             time_axis = np.linspace(0, len(audio) / self.sr, len(audio))
             ax.plot(time_axis, audio, linewidth=0.5, alpha=0.8)
@@ -64,30 +67,33 @@ class AudioVisualizer:
         else:
             ax.plot(audio, linewidth=0.5, alpha=0.8)
             ax.set_xlabel('样本点')
-        
+
         ax.set_ylabel('幅度')
         ax.set_title(title)
         ax.grid(True, alpha=0.3)
-        
+
         # 添加零线
         ax.axhline(y=0, color='red', linestyle='--', alpha=0.5)
-        
+
         plt.tight_layout()
         return fig
-    
-    def plot_spectrogram(self, audio: np.ndarray, title: str = "频谱图", 
-                        n_fft: int = 2048, hop_length: int = 512,
-                        ax: Optional[plt.Axes] = None) -> plt.Figure:
+
+    def plot_spectrogram(self,
+                         audio: np.ndarray,
+                         title: str = "频谱图",
+                         n_fft: int = 2048,
+                         hop_length: int = 512,
+                         ax: Optional[plt.Axes] = None) -> plt.Figure:
         """
         绘制频谱图
-        
+
         Args:
             audio: 音频信号
             title: 图标题
             n_fft: FFT窗口大小
             hop_length: 跳跃长度
             ax: matplotlib轴对象
-            
+
         Returns:
             图形对象
         """
@@ -95,37 +101,44 @@ class AudioVisualizer:
             fig, ax = plt.subplots(figsize=self.figsize)
         else:
             fig = ax.figure
-        
+
         # 计算频谱图
         D = librosa.stft(audio, n_fft=n_fft, hop_length=hop_length)
         S_db = librosa.amplitude_to_db(np.abs(D), ref=np.max)
-        
+
         # 绘制
-        img = librosa.display.specshow(S_db, sr=self.sr, hop_length=hop_length,
-                                      x_axis='time', y_axis='hz', ax=ax)
-        
+        img = librosa.display.specshow(S_db,
+                                       sr=self.sr,
+                                       hop_length=hop_length,
+                                       x_axis='time',
+                                       y_axis='hz',
+                                       ax=ax)
+
         ax.set_title(title)
         ax.set_xlabel('时间 (s)')
         ax.set_ylabel('频率 (Hz)')
-        
+
         # 添加颜色条
         cbar = plt.colorbar(img, ax=ax, format='%+2.0f dB')
         cbar.set_label('幅度 (dB)')
-        
+
         plt.tight_layout()
         return fig
-    
-    def plot_mel_spectrogram(self, audio: np.ndarray, title: str = "梅尔频谱图",
-                            n_mels: int = 128, ax: Optional[plt.Axes] = None) -> plt.Figure:
+
+    def plot_mel_spectrogram(self,
+                             audio: np.ndarray,
+                             title: str = "梅尔频谱图",
+                             n_mels: int = 128,
+                             ax: Optional[plt.Axes] = None) -> plt.Figure:
         """
         绘制梅尔频谱图
-        
+
         Args:
             audio: 音频信号
             title: 图标题
             n_mels: 梅尔滤波器数量
             ax: matplotlib轴对象
-            
+
         Returns:
             图形对象
         """
@@ -133,37 +146,43 @@ class AudioVisualizer:
             fig, ax = plt.subplots(figsize=self.figsize)
         else:
             fig = ax.figure
-        
+
         # 计算梅尔频谱图
         S = librosa.feature.melspectrogram(y=audio, sr=self.sr, n_mels=n_mels)
         S_db = librosa.power_to_db(S, ref=np.max)
-        
+
         # 绘制
-        img = librosa.display.specshow(S_db, sr=self.sr, x_axis='time', 
-                                      y_axis='mel', ax=ax)
-        
+        img = librosa.display.specshow(S_db,
+                                       sr=self.sr,
+                                       x_axis='time',
+                                       y_axis='mel',
+                                       ax=ax)
+
         ax.set_title(title)
         ax.set_xlabel('时间 (s)')
         ax.set_ylabel('梅尔频率')
-        
+
         # 添加颜色条
         cbar = plt.colorbar(img, ax=ax, format='%+2.0f dB')
         cbar.set_label('功率 (dB)')
-        
+
         plt.tight_layout()
         return fig
-    
-    def plot_spectrum(self, audio: np.ndarray, title: str = "频谱", 
-                     log_scale: bool = True, ax: Optional[plt.Axes] = None) -> plt.Figure:
+
+    def plot_spectrum(self,
+                      audio: np.ndarray,
+                      title: str = "频谱",
+                      log_scale: bool = True,
+                      ax: Optional[plt.Axes] = None) -> plt.Figure:
         """
         绘制频谱
-        
+
         Args:
             audio: 音频信号
             title: 图标题
             log_scale: 是否使用对数刻度
             ax: matplotlib轴对象
-            
+
         Returns:
             图形对象
         """
@@ -171,17 +190,17 @@ class AudioVisualizer:
             fig, ax = plt.subplots(figsize=self.figsize)
         else:
             fig = ax.figure
-        
+
         # 计算FFT
         fft_data = np.fft.fft(audio)
         magnitude = np.abs(fft_data)
-        freqs = np.fft.fftfreq(len(audio), 1/self.sr)
-        
+        freqs = np.fft.fftfreq(len(audio), 1 / self.sr)
+
         # 只取正频率部分
         positive_idx = freqs >= 0
         freqs = freqs[positive_idx]
         magnitude = magnitude[positive_idx]
-        
+
         if log_scale:
             magnitude_db = 20 * np.log10(magnitude + 1e-10)
             ax.plot(freqs, magnitude_db)
@@ -189,280 +208,325 @@ class AudioVisualizer:
         else:
             ax.plot(freqs, magnitude)
             ax.set_ylabel('幅度')
-        
+
         ax.set_xlabel('频率 (Hz)')
         ax.set_title(title)
         ax.grid(True, alpha=0.3)
-        
+
         plt.tight_layout()
         return fig
-    
-    def plot_features_comparison(self, features_dict: Dict[str, np.ndarray], 
-                               title: str = "特征对比") -> plt.Figure:
+
+    def plot_features_comparison(self,
+                                 features_dict: Dict[str, np.ndarray],
+                                 title: str = "特征对比") -> plt.Figure:
         """
         绘制多个特征的对比图
-        
+
         Args:
             features_dict: 特征字典 {特征名: 特征值数组}
             title: 图标题
-            
+
         Returns:
             图形对象
         """
         n_features = len(features_dict)
-        fig, axes = plt.subplots(n_features, 1, figsize=(self.figsize[0], self.figsize[1] * n_features / 2))
-        
+        fig, axes = plt.subplots(n_features,
+                                 1,
+                                 figsize=(self.figsize[0],
+                                          self.figsize[1] * n_features / 2))
+
         if n_features == 1:
             axes = [axes]
-        
-        for i, (feature_name, feature_values) in enumerate(features_dict.items()):
+
+        for i, (feature_name,
+                feature_values) in enumerate(features_dict.items()):
             if len(feature_values.shape) == 1:
                 # 一维特征
-                time_axis = np.linspace(0, len(feature_values) / (self.sr / 512), len(feature_values))
+                time_axis = np.linspace(0,
+                                        len(feature_values) / (self.sr / 512),
+                                        len(feature_values))
                 axes[i].plot(time_axis, feature_values)
                 axes[i].set_ylabel(feature_name)
             else:
                 # 二维特征（如MFCC）
-                img = axes[i].imshow(feature_values, aspect='auto', origin='lower')
+                img = axes[i].imshow(feature_values,
+                                     aspect='auto',
+                                     origin='lower')
                 axes[i].set_ylabel(feature_name)
                 plt.colorbar(img, ax=axes[i])
-            
+
             axes[i].set_title(f'{feature_name} 特征')
             axes[i].grid(True, alpha=0.3)
-        
+
         axes[-1].set_xlabel('时间 (s)')
         plt.suptitle(title)
         plt.tight_layout()
         return fig
-    
-    def plot_statistics_distribution(self, stats_dict: Dict[str, List[float]], 
-                                   title: str = "统计分布图") -> plt.Figure:
+
+    def plot_statistics_distribution(self,
+                                     stats_dict: Dict[str, List[float]],
+                                     title: str = "统计分布图") -> plt.Figure:
         """
         绘制统计分布图
-        
+
         Args:
             stats_dict: 统计数据字典
             title: 图标题
-            
+
         Returns:
             图形对象
         """
         n_stats = len(stats_dict)
-        fig, axes = plt.subplots(2, (n_stats + 1) // 2, figsize=(self.figsize[0], self.figsize[1]))
-        
+        fig, axes = plt.subplots(2, (n_stats + 1) // 2,
+                                 figsize=(self.figsize[0], self.figsize[1]))
+
         if n_stats == 1:
             axes = [axes]
         elif n_stats == 2:
             axes = axes.flatten()
         else:
             axes = axes.flatten()
-        
+
         for i, (stat_name, values) in enumerate(stats_dict.items()):
             if i >= len(axes):
                 break
-            
+
             # 绘制直方图和KDE
-            axes[i].hist(values, bins=30, alpha=0.7, density=True, color='skyblue')
-            
+            axes[i].hist(values,
+                         bins=30,
+                         alpha=0.7,
+                         density=True,
+                         color='skyblue')
+
             try:
                 sns.kdeplot(values, ax=axes[i], color='red')
             except:
                 pass
-            
+
             axes[i].set_title(f'{stat_name} 分布')
             axes[i].set_xlabel(stat_name)
             axes[i].set_ylabel('密度')
             axes[i].grid(True, alpha=0.3)
-        
+
         # 隐藏未使用的子图
         for j in range(i + 1, len(axes)):
             axes[j].set_visible(False)
-        
+
         plt.suptitle(title)
         plt.tight_layout()
         return fig
-    
-    def plot_rms_distribution(self, rms_values: List[float], 
-                             title: str = "RMS分布图") -> plt.Figure:
+
+    def plot_rms_distribution(self,
+                              rms_values: List[float],
+                              title: str = "RMS分布图") -> plt.Figure:
         """
         绘制RMS分布图
-        
+
         Args:
             rms_values: RMS值列表
             title: 图标题
-            
+
         Returns:
             图形对象
         """
         fig, (ax1, ax2) = plt.subplots(1, 2, figsize=self.figsize)
-        
+
         # 线性尺度分布
-        ax1.hist(rms_values, bins=50, alpha=0.7, color='lightblue', edgecolor='black')
+        ax1.hist(rms_values,
+                 bins=50,
+                 alpha=0.7,
+                 color='lightblue',
+                 edgecolor='black')
         ax1.set_xlabel('RMS 幅度')
         ax1.set_ylabel('频次')
         ax1.set_title('RMS 线性分布')
         ax1.grid(True, alpha=0.3)
-        
+
         # 对数尺度分布
         rms_db = [20 * np.log10(rms + 1e-10) for rms in rms_values]
-        ax2.hist(rms_db, bins=50, alpha=0.7, color='lightgreen', edgecolor='black')
+        ax2.hist(rms_db,
+                 bins=50,
+                 alpha=0.7,
+                 color='lightgreen',
+                 edgecolor='black')
         ax2.set_xlabel('RMS (dB)')
         ax2.set_ylabel('频次')
         ax2.set_title('RMS 对数分布')
         ax2.grid(True, alpha=0.3)
-        
+
         plt.suptitle(title)
         plt.tight_layout()
         return fig
-    
-    def plot_audio_comparison(self, audio1: np.ndarray, audio2: np.ndarray,
-                             labels: List[str] = None, title: str = "音频对比") -> plt.Figure:
+
+    def plot_audio_comparison(self,
+                              audio1: np.ndarray,
+                              audio2: np.ndarray,
+                              labels: List[str] = None,
+                              title: str = "音频对比") -> plt.Figure:
         """
         绘制两个音频的对比图
-        
+
         Args:
             audio1: 第一个音频
             audio2: 第二个音频
             labels: 标签列表
             title: 图标题
-            
+
         Returns:
             图形对象
         """
         if labels is None:
             labels = ['音频1', '音频2']
-        
-        fig, axes = plt.subplots(3, 2, figsize=(self.figsize[0], self.figsize[1] * 1.5))
-        
+
+        fig, axes = plt.subplots(3,
+                                 2,
+                                 figsize=(self.figsize[0],
+                                          self.figsize[1] * 1.5))
+
         # 时域波形对比
         time1 = np.linspace(0, len(audio1) / self.sr, len(audio1))
         time2 = np.linspace(0, len(audio2) / self.sr, len(audio2))
-        
+
         axes[0, 0].plot(time1, audio1, alpha=0.8)
         axes[0, 0].set_title(f'{labels[0]} - 波形')
         axes[0, 0].set_xlabel('时间 (s)')
         axes[0, 0].set_ylabel('幅度')
         axes[0, 0].grid(True, alpha=0.3)
-        
+
         axes[0, 1].plot(time2, audio2, alpha=0.8, color='orange')
         axes[0, 1].set_title(f'{labels[1]} - 波形')
         axes[0, 1].set_xlabel('时间 (s)')
         axes[0, 1].set_ylabel('幅度')
         axes[0, 1].grid(True, alpha=0.3)
-        
+
         # 频谱对比
         self.plot_spectrum(audio1, f'{labels[0]} - 频谱', ax=axes[1, 0])
         self.plot_spectrum(audio2, f'{labels[1]} - 频谱', ax=axes[1, 1])
-        
+
         # 频谱图对比
         self.plot_spectrogram(audio1, f'{labels[0]} - 频谱图', ax=axes[2, 0])
         self.plot_spectrogram(audio2, f'{labels[1]} - 频谱图', ax=axes[2, 1])
-        
+
         plt.suptitle(title)
         plt.tight_layout()
         return fig
 
 
-def plot_dataset_overview(file_paths: List[str], max_files: int = 10, 
-                         sr: int = 22050) -> plt.Figure:
+def plot_dataset_overview(file_paths: List[str],
+                          max_files: int = 10,
+                          sr: int = 22050) -> plt.Figure:
     """
     绘制数据集概览
-    
+
     Args:
         file_paths: 音频文件路径列表
         max_files: 最大显示文件数
         sr: 采样率
-        
+
     Returns:
         图形对象
     """
     visualizer = AudioVisualizer(sr=sr)
-    
+
     # 限制文件数量
     selected_files = file_paths[:max_files]
-    
-    fig, axes = plt.subplots(len(selected_files), 2, 
-                            figsize=(15, 3 * len(selected_files)))
-    
+
+    fig, axes = plt.subplots(len(selected_files),
+                             2,
+                             figsize=(15, 3 * len(selected_files)))
+
     if len(selected_files) == 1:
         axes = axes.reshape(1, -1)
-    
+
     for i, file_path in enumerate(selected_files):
         try:
             audio, _ = librosa.load(file_path, sr=sr)
-            
+
             # 波形图
-            visualizer.plot_waveform(audio, f'文件 {i+1}: 波形', ax=axes[i, 0])
-            
+            visualizer.plot_waveform(audio, f'文件 {i + 1}: 波形', ax=axes[i, 0])
+
             # 频谱图
-            visualizer.plot_spectrogram(audio, f'文件 {i+1}: 频谱图', ax=axes[i, 1])
-            
+            visualizer.plot_spectrogram(audio, f'文件 {i + 1}: 频谱图', ax=axes[i, 1])
+
         except Exception as e:
-            axes[i, 0].text(0.5, 0.5, f'加载失败: {str(e)}', 
-                           ha='center', va='center', transform=axes[i, 0].transAxes)
-            axes[i, 1].text(0.5, 0.5, f'加载失败: {str(e)}', 
-                           ha='center', va='center', transform=axes[i, 1].transAxes)
-    
+            axes[i, 0].text(0.5,
+                            0.5,
+                            f'加载失败: {str(e)}',
+                            ha='center',
+                            va='center',
+                            transform=axes[i, 0].transAxes)
+            axes[i, 1].text(0.5,
+                            0.5,
+                            f'加载失败: {str(e)}',
+                            ha='center',
+                            va='center',
+                            transform=axes[i, 1].transAxes)
+
     plt.suptitle('数据集概览')
     plt.tight_layout()
     return fig
 
 
-def create_analysis_dashboard(audio: np.ndarray, sr: int = 22050) -> plt.Figure:
+def create_analysis_dashboard(audio: np.ndarray,
+                              sr: int = 22050) -> plt.Figure:
     """
     创建音频分析仪表板
-    
+
     Args:
         audio: 音频信号
         sr: 采样率
-        
+
     Returns:
         仪表板图形对象
     """
     visualizer = AudioVisualizer(sr=sr)
-    
+
     fig = plt.figure(figsize=(16, 12))
-    
+
     # 创建网格布局
     gs = fig.add_gridspec(3, 3, hspace=0.3, wspace=0.3)
-    
+
     # 时域波形
     ax1 = fig.add_subplot(gs[0, :])
     visualizer.plot_waveform(audio, "时域波形", ax=ax1)
-    
+
     # 频谱图
     ax2 = fig.add_subplot(gs[1, :2])
     visualizer.plot_spectrogram(audio, "频谱图", ax=ax2)
-    
+
     # 频谱
     ax3 = fig.add_subplot(gs[1, 2])
     visualizer.plot_spectrum(audio, "频谱", ax=ax3)
-    
+
     # 梅尔频谱图
     ax4 = fig.add_subplot(gs[2, :2])
     visualizer.plot_mel_spectrogram(audio, "梅尔频谱图", ax=ax4)
-    
+
     # 特征统计
     ax5 = fig.add_subplot(gs[2, 2])
-    
+
     # 计算基本统计
     duration = len(audio) / sr
     max_amp = np.max(np.abs(audio))
-    rms_amp = np.sqrt(np.mean(audio ** 2))
-    
+    rms_amp = np.sqrt(np.mean(audio**2))
+
     stats_text = f"""音频统计信息:
-时长: {duration:.2f}s
-最大幅度: {max_amp:.4f}
-RMS: {rms_amp:.4f}
-RMS (dB): {20*np.log10(rms_amp):.2f}
-采样率: {sr} Hz
-样本数: {len(audio)}
-"""
-    
-    ax5.text(0.1, 0.5, stats_text, transform=ax5.transAxes, 
-             fontsize=10, verticalalignment='center')
+                        时长: {duration:.2f}s
+                        最大幅度: {max_amp:.4f}
+                        RMS: {rms_amp:.4f}
+                        RMS (dB): {20 * np.log10(rms_amp):.2f}
+                        采样率: {sr} Hz
+                        样本数: {len(audio)}
+                        """
+
+    ax5.text(0.1,
+             0.5,
+             stats_text,
+             transform=ax5.transAxes,
+             fontsize=10,
+             verticalalignment='center')
     ax5.set_title("统计信息")
     ax5.axis('off')
-    
+
     plt.suptitle("音频分析仪表板", fontsize=16)
-    return fig
+    return fig