mirror of
https://github.com/xinnan-tech/xiaozhi-esp32-server.git
synced 2026-07-29 03:53:55 +08:00
huozaimengli
15650e1a6c
重构所有ASR提供商的speech_to_text方法，将重复的音频解码、合并和文件保存逻辑提取到基类的speech_to_text_wrapper中。引入AudioArtifacts数据类封装PCM帧、字节数据、文件路径和临时路径，简化各提供商实现。移除各提供商中的冗余文件清理代码，由基类统一处理。 新增requires_file()和prefers_temp_file()方法允许提供商声明文件需求，优化内存和磁盘使用。保持接口兼容性的同时提高代码复用性和可维护性。
93 lines
3.3 KiB
Python
93 lines
3.3 KiB
Python
import os
|
|
import json
|
|
import time
|
|
from typing import Optional, Tuple, List
|
|
from .base import ASRProviderBase
|
|
from config.logger import setup_logging
|
|
from core.providers.asr.dto.dto import InterfaceType
|
|
import vosk
|
|
|
|
TAG = __name__
|
|
logger = setup_logging()
|
|
|
|
class ASRProvider(ASRProviderBase):
|
|
def __init__(self, config: dict, delete_audio_file: bool = True):
|
|
super().__init__()
|
|
self.interface_type = InterfaceType.LOCAL
|
|
self.model_path = config.get("model_path")
|
|
self.output_dir = config.get("output_dir", "tmp/")
|
|
self.delete_audio_file = delete_audio_file
|
|
|
|
# 初始化VOSK模型
|
|
self.model = None
|
|
self.recognizer = None
|
|
self._load_model()
|
|
|
|
# 确保输出目录存在
|
|
os.makedirs(self.output_dir, exist_ok=True)
|
|
|
|
def _load_model(self):
|
|
"""加载VOSK模型"""
|
|
try:
|
|
if not os.path.exists(self.model_path):
|
|
raise FileNotFoundError(f"VOSK模型路径不存在: {self.model_path}")
|
|
|
|
logger.bind(tag=TAG).info(f"正在加载VOSK模型: {self.model_path}")
|
|
self.model = vosk.Model(self.model_path)
|
|
|
|
# 初始化VOSK识别器（采样率必须为16kHz）
|
|
self.recognizer = vosk.KaldiRecognizer(self.model, 16000)
|
|
|
|
logger.bind(tag=TAG).info("VOSK模型加载成功")
|
|
except Exception as e:
|
|
logger.bind(tag=TAG).error(f"加载VOSK模型失败: {e}")
|
|
raise
|
|
|
|
async def speech_to_text(
|
|
self, opus_data: List[bytes], session_id: str, audio_format="opus"
|
|
) -> Tuple[Optional[str], Optional[str]]:
|
|
"""将语音数据转换为文本"""
|
|
try:
|
|
# 检查模型是否加载成功
|
|
if not self.model:
|
|
logger.bind(tag=TAG).error("VOSK模型未加载，无法进行识别")
|
|
return "", None
|
|
|
|
artifacts = self.get_current_artifacts()
|
|
if artifacts is None:
|
|
return "", None
|
|
if not artifacts.pcm_bytes:
|
|
logger.bind(tag=TAG).warning("合并后的PCM数据为空")
|
|
return "", None
|
|
|
|
start_time = time.time()
|
|
|
|
|
|
# 进行识别（VOSK推荐每次送入2000字节的数据）
|
|
chunk_size = 2000
|
|
text_result = ""
|
|
|
|
for i in range(0, len(artifacts.pcm_bytes), chunk_size):
|
|
chunk = artifacts.pcm_bytes[i:i+chunk_size]
|
|
if self.recognizer.AcceptWaveform(chunk):
|
|
result = json.loads(self.recognizer.Result())
|
|
text = result.get('text', '')
|
|
if text:
|
|
text_result += text + " "
|
|
|
|
# 获取最终结果
|
|
final_result = json.loads(self.recognizer.FinalResult())
|
|
final_text = final_result.get('text', '')
|
|
if final_text:
|
|
text_result += final_text
|
|
|
|
logger.bind(tag=TAG).debug(
|
|
f"VOSK语音识别耗时: {time.time() - start_time:.3f}s | 结果: {text_result.strip()}"
|
|
)
|
|
|
|
return text_result.strip(), artifacts.file_path
|
|
|
|
except Exception as e:
|
|
logger.bind(tag=TAG).error(f"VOSK语音识别失败: {e}")
|
|
return "", None
|