File size: 7,938 Bytes
64f8498 25666e3 64f8498 25666e3 64f8498 25666e3 64f8498 25666e3 64f8498 25666e3 64f8498 25666e3 64f8498 7fc9887 64f8498 25666e3 64f8498 25666e3 64f8498 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 |
from faster_whisper import WhisperModel
from audio_processor import AudioSegment
import json
from typing import List, Dict, Optional, Tuple
from dataclasses import dataclass
import os
from datetime import datetime
import numpy as np
@dataclass
class TranscriptionResult:
text: str
start_time: float
end_time: float
confidence: float
verified: bool = False
verified_text: Optional[str] = None
verification_notes: Optional[str] = None
segment_index: Optional[int] = None # 添加片段索引字段
class AudioTranscriber:
def __init__(self, model: str = "large", device: str = "cuda", compute_type: str = "int8"):
"""
初始化转录器
Args:
model_path: Whisper模型路径
device: 使用的设备 ("cpu" 或 "cuda")
compute_type: 计算类型
"""
print("📥 Loading Whisper model...")
self.model = WhisperModel(model, device=device, compute_type=compute_type)
print("📥 Loading Whisper model successfully!!")
def transcribe_segment(self, segment: AudioSegment) -> List[TranscriptionResult]:
"""
转录单个音频片段
"""
print("Model transcribe...")
print(f"开始转录音频片段，长度: {len(segment.audio_data)} 采样点 ({len(segment.audio_data)/16000:.2f}秒)")
segments_generator, info = self.model.transcribe(segment.audio_data,
beam_size=5,
language="zh")
print(f"检测到语言: {info.language}, 语言概率: {info.language_probability:.2f}")
segments = list(segments_generator)
print(f"Model transcribe successfully! Segments count: {len(segments)}")
if len(segments) > 0:
print(segments[0])
results = []
for seg in segments:
# 调整时间戳以匹配原始音频中的位置
absolute_start = segment.start_time + seg.start
absolute_end = segment.start_time + seg.end
result = TranscriptionResult(
text=seg.text,
start_time=absolute_start,
end_time=absolute_end,
confidence=1.0 - seg.no_speech_prob
)
results.append(result)
return results
def transcribe_aggregated_segments(self, segments: List[AudioSegment]) -> List[TranscriptionResult]:
"""
将多个连续的音频片段聚合后进行转录，以提高语义准确度
Args:
segments: 要聚合的连续音频片段列表
Returns:
转录结果列表
"""
if not segments:
return []
print(f"开始聚合转录 {len(segments)} 个连续片段...")
# 记录第一个片段的开始时间和最后一个片段的结束时间
start_time = segments[0].start_time
end_time = segments[-1].end_time
# 计算总长度并创建合并的音频数据数组
total_length = sum(len(segment.audio_data) for segment in segments)
aggregated_audio = np.zeros(total_length, dtype=np.float32)
# 合并音频数据
current_position = 0
for segment in segments:
segment_length = len(segment.audio_data)
aggregated_audio[current_position:current_position + segment_length] = segment.audio_data
current_position += segment_length
print(f"聚合后音频长度: {len(aggregated_audio)} 采样点 ({len(aggregated_audio)/16000:.2f}秒)")
# 创建一个临时的聚合片段对象
aggregated_segment = AudioSegment(
start_time=start_time,
end_time=end_time,
audio_data=aggregated_audio,
is_speech=True
)
# 转录聚合后的音频
print("开始转录聚合后的音频...")
return self.transcribe_segment(aggregated_segment)
def save_transcription(self,
results: List[TranscriptionResult],
audio_path: str,
output_dir: str = "dataset/transcripts"):
"""
保存转录结果到JSON文件
"""
# 生成输出文件名
base_name = os.path.splitext(os.path.basename(audio_path))[0]
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
output_path = os.path.join(output_dir, f"{base_name}_{timestamp}.json")
# 准备保存的数据
data = {
"audio_file": audio_path,
"timestamp": timestamp,
"segments": [
{
"text": r.text,
"start_time": r.start_time,
"end_time": r.end_time,
"confidence": r.confidence,
"verified": r.verified,
"verified_text": r.verified_text,
"verification_notes": r.verification_notes,
"segment_index": r.segment_index # 添加片段索引到输出
}
for r in results
]
}
# 保存到文件
os.makedirs(output_dir, exist_ok=True)
with open(output_path, 'w', encoding='utf-8') as f:
json.dump(data, f, ensure_ascii=False, indent=2)
return output_path
def verify_transcription(self,
result: TranscriptionResult,
verified_text: Optional[str] = None,
verification_notes: Optional[str] = None) -> TranscriptionResult:
"""
人工验证转录结果
"""
result.verified = True
if verified_text is not None:
result.verified_text = verified_text
if verification_notes is not None:
result.verification_notes = verification_notes
return result
def load_transcription(self, json_path: str) -> List[TranscriptionResult]:
"""
从JSON文件加载转录结果
"""
with open(json_path, 'r', encoding='utf-8') as f:
data = json.load(f)
results = []
for seg in data["segments"]:
result = TranscriptionResult(
text=seg["text"],
start_time=seg["start_time"],
end_time=seg["end_time"],
confidence=seg["confidence"],
verified=seg["verified"],
verified_text=seg.get("verified_text"),
verification_notes=seg.get("verification_notes"),
segment_index=seg.get("segment_index") # 加载片段索引
)
results.append(result)
return results
if __name__ == "__main__":
# 测试代码
from audio_processor import AudioProcessor
# 初始化处理器和转录器
processor = AudioProcessor()
transcriber = AudioTranscriber()
# 示例：处理和转录音频文件
audio_path = "../dataset/audio/test.wav" # 替换为实际的音频文件路径
try:
# 1. 切割音频
segments = processor.process_audio_file(audio_path)
print(f"检测到 {len(segments)} 个语音片段")
# 2. 转录每个片段
all_results = []
for i, segment in enumerate(segments):
print(f"转录片段 {i+1}/{len(segments)}...")
results = transcriber.transcribe_segment(segment)
all_results.extend(results)
# 3. 保存结果
output_path = transcriber.save_transcription(all_results, audio_path)
print(f"✅ 转录结果已保存到: {output_path}")
except Exception as e:
print(f"处理音频时出错: {e}")
|