Update app.py

app.py

@@ -86,29 +86,27 @@ def calculate_overlap(start1, end1, start2, end2):
     overlap_duration = max(0, overlap_end - overlap_start)
     return overlap_duration
 
-# 获取目标时间段和说话人时间段的重叠比例
-def get_matching_segments(target_time, diarization_output):
-    target_start_time = target_time['start_time']
-    target_end_time = target_time['end_time']
-    
-    # 获取该说话人时间段的信息，排除目标录音时间段
+# 获取所有说话人时间段（排除目标录音时间段）
+def get_all_speaker_segments(diarization_output, target_start_time, target_end_time, final_audio_length):
     speaker_segments = {}
-    for speech_turn in diarization_output.itertracks(yield_label=True):  # 使用 itertracks 获取每个说话人的信息
+    
+    # 使用 itertracks 获取每个说话人的信息
+    for speech_turn in diarization_output.itertracks(yield_label=True):
         start_seconds = speech_turn[0].start
         end_seconds = speech_turn[0].end
         label = speech_turn[1]
         
-        # 计算目标音频时间段和说话人时间段的重叠时间
-        overlap = calculate_overlap(target_start_time, target_end_time, start_seconds, end_seconds)
+        # 检查时间段是否与目标录音重叠
+        if start_seconds < target_end_time and end_seconds > target_start_time:
+            # 如果时间段与目标音频有重叠，调整结束时间
+            end_seconds = min(end_seconds, final_audio_length)
         
-        # 如果存在重叠，排除目标音频时间段
-        if overlap > 0:
-            if label not in speaker_segments:
-                speaker_segments[label] = []
-            
-            # 如果时间段与目标音频有重叠，跳过该时间段
-            if start_seconds >= target_end_time or end_seconds <= target_start_time:
-                speaker_segments[label].append((start_seconds, end_seconds))
+        # 存储说话人的时间段
+        if label not in speaker_segments:
+            speaker_segments[label] = []
+        
+        # 添加处理后的时间段
+        speaker_segments[label].append((start_seconds, end_seconds))
 
     return speaker_segments
 
@@ -126,20 +124,23 @@ def process_audio(target_audio, mixed_audio):
     if isinstance(diarization_result, str) and diarization_result.startswith("错误"):
         return diarization_result, None  # 出错时返回错误信息
     else:
-        # 获取该说话人的所有匹配时间段（排除目标音频时间段）
-        matching_segments = get_matching_segments(time_dict, diarization_result)
+        # 获取拼接后的音频长度
+        final_audio_length = len(AudioSegment.from_wav("final_output.wav")) / 1000  # 秒为单位
+        
+        # 获取所有说话人的时间段
+        speaker_segments = get_all_speaker_segments(diarization_result, time_dict['start_time'], time_dict['end_time'], final_audio_length)
         
-        if matching_segments:
-            # 返回匹配的说话人标签和他们的时间段
-            return matching_segments
+        if speaker_segments:
+            # 返回所有说话人的时间段
+            return speaker_segments
         else:
-            return "没有找到匹配的说话人时间段。"
+            return "没有找到任何说话人的时间段。"
 
 # Gradio 接口
 with gr.Blocks() as demo:
     gr.Markdown("""
     # 🗣️ 音频拼接与说话人分类 🗣️
-    上传目标音频和混合音频，拼接并进行说话人分类。结果包括所有匹配说话人的时间段（排除目标录音时间段）。
+    上传目标音频和混合音频，拼接并进行说话人分类。结果包括所有说话人的时间段（排除目标录音时间段）。
     """)
     
     mixed_audio_input = gr.Audio(type="filepath", label="上传混合音频")
@@ -148,7 +149,7 @@ with gr.Blocks() as demo:
     process_button = gr.Button("处理音频")
     
     # 输出结果
-    diarization_output = gr.Textbox(label="匹配的说话人时间段")
+    diarization_output = gr.Textbox(label="说话人时间段")
 
     # 点击按钮时触发处理音频
     process_button.click(