226 lines
10 KiB
Python
226 lines
10 KiB
Python
import sounddevice as sd
|
||
import pvporcupine
|
||
import struct
|
||
import websocket
|
||
import threading
|
||
import hmac
|
||
import hashlib
|
||
import base64
|
||
import json
|
||
import time
|
||
import urllib.parse
|
||
import uuid
|
||
import queue
|
||
import sys
|
||
|
||
# 原代码9. 音频采集与WebSocket + 10. 唤醒词监听完整逻辑
|
||
class VoiceRecogController:
|
||
def __init__(self, access_key, wakeup_word_path, model_path, appid, access_key_id, access_key_secret, tts_controller, feedback_text):
|
||
# 接收调度脚本传入的参数，保持原逻辑
|
||
self.ACCESS_KEY = access_key
|
||
self.WAKEUP_WORD_PATH = wakeup_word_path
|
||
self.MODEL_PATH = model_path
|
||
self.APPID = appid
|
||
self.ACCESS_KEY_ID = access_key_id
|
||
self.ACCESS_KEY_SECRET = access_key_secret
|
||
self.tts_controller = tts_controller
|
||
self.FEEDBACK_TEXT = feedback_text
|
||
self.SAMPLE_RATE = 16000
|
||
self.CHANNELS = 1
|
||
self.SAMPLE_FORMAT = "int16"
|
||
self.INTERACTION_TIMEOUT = 30
|
||
self.audio_q = queue.Queue()
|
||
self.stream = None # 麦克风流后续初始化
|
||
|
||
def wakeup_listener(self):
|
||
"""原代码10. 唤醒词监听"""
|
||
try:
|
||
porcupine = pvporcupine.create(
|
||
access_key=self.ACCESS_KEY,
|
||
keyword_paths=[self.WAKEUP_WORD_PATH],
|
||
model_path=self.MODEL_PATH
|
||
)
|
||
print(f"\n🎯 唤醒词引擎就绪（采样率：{porcupine.sample_rate}）")
|
||
|
||
wakeup_mic = sd.RawInputStream(
|
||
samplerate=porcupine.sample_rate,
|
||
blocksize=porcupine.frame_length,
|
||
dtype="int16",
|
||
channels=1
|
||
)
|
||
|
||
print("📢 等待唤醒词「小黄鸭」（按Ctrl+C退出）")
|
||
with wakeup_mic:
|
||
while True:
|
||
pcm_data, _ = wakeup_mic.read(porcupine.frame_length)
|
||
pcm_unpacked = struct.unpack_from("h" * porcupine.frame_length, pcm_data)
|
||
if porcupine.process(pcm_unpacked) >= 0:
|
||
print("🚀 检测到唤醒词「小黄鸭」！")
|
||
# 播放唤醒反馈（同步执行）
|
||
self.tts_controller.speak(self.FEEDBACK_TEXT["wakeup"])
|
||
porcupine.delete()
|
||
return True
|
||
except Exception as e:
|
||
print(f"\n❌ 唤醒词监听失败：{str(e)}")
|
||
print(" 排查：1. 唤醒词文件路径 2. 麦克风连接 3. PicoVoice Key有效性")
|
||
sys.exit(1)
|
||
|
||
def _audio_callback(self, indata, frames, t, status):
|
||
"""原代码9. 音频采集回调"""
|
||
if status:
|
||
print(f"⚠️ 音频异常：{status}")
|
||
self.audio_q.put(bytes(indata))
|
||
|
||
def _create_ws_url(self):
|
||
"""原代码9. 创建WebSocket URL"""
|
||
try:
|
||
host = "office-api-ast-dx.iflyaisol.com"
|
||
path = "/ast/communicate/v1"
|
||
utc = time.strftime("%Y-%m-%dT%H:%M:%S", time.gmtime()) + "+0000"
|
||
session_uuid = str(uuid.uuid4())
|
||
|
||
params = {
|
||
"accessKeyId": self.ACCESS_KEY_ID,
|
||
"appId": self.APPID,
|
||
"samplerate": self.SAMPLE_RATE,
|
||
"audio_encode": "pcm_s16le",
|
||
"lang": "autodialect",
|
||
"uuid": session_uuid,
|
||
"utc": utc,
|
||
}
|
||
|
||
sorted_params = sorted(params.items(), key=lambda x: x[0])
|
||
base_string = "&".join(
|
||
f"{urllib.parse.quote_plus(str(k))}={urllib.parse.quote_plus(str(v))}"
|
||
for k, v in sorted_params
|
||
)
|
||
signature = hmac.new(
|
||
self.ACCESS_KEY_SECRET.encode("utf-8"),
|
||
base_string.encode("utf-8"),
|
||
hashlib.sha1
|
||
).digest()
|
||
signature = base64.b64encode(signature).decode("utf-8")
|
||
|
||
query = base_string + "&signature=" + urllib.parse.quote_plus(signature)
|
||
return f"wss://{host}{path}?{query}", session_uuid
|
||
except Exception as e:
|
||
print(f"❌ WebSocket URL生成失败：{str(e)}")
|
||
return None, None
|
||
|
||
def _on_message(self, ws, message, current_text, last_audio_time):
|
||
"""原代码9. WebSocket消息处理（接收全局变量引用）"""
|
||
try:
|
||
data = json.loads(message)
|
||
if data.get("msg_type") == "result" and "cn" in data.get("data", {}):
|
||
words = [
|
||
cw.get("w", "")
|
||
for rt in data["data"]["cn"].get("st", {}).get("rt", [])
|
||
for ws_item in rt.get("ws", [])
|
||
for cw in ws_item.get("cw", [])
|
||
]
|
||
if words:
|
||
current_text[0] = "".join(words) # 用列表传引用，修改全局变量
|
||
last_audio_time[0] = time.time()
|
||
print(f"🎧 识别中：{current_text[0]}", end="\r")
|
||
except Exception as e:
|
||
print(f"\n❌ 语音识别消息处理错误：{str(e)}")
|
||
|
||
def _on_error(self, ws, error, is_processing):
|
||
"""原代码9. WebSocket错误处理"""
|
||
if not is_processing[0]:
|
||
print(f"\n❌ WebSocket连接错误：{str(error)}")
|
||
|
||
def _on_close(self, ws, close_status_code, close_msg, current_text, final_result, stream):
|
||
"""原代码9. WebSocket关闭处理"""
|
||
print(f"\n🔌 WebSocket连接关闭 | 状态码：{close_status_code}")
|
||
if stream and stream.active:
|
||
stream.stop()
|
||
current_text[0] = ""
|
||
final_result[0] = ""
|
||
|
||
def _on_open(self, ws, stream, current_text, final_result, last_audio_time, is_processing, last_command_time, execute_callback):
|
||
"""新增 execute_callback 参数，用于接收指令执行函数"""
|
||
def send_audio_and_handle():
|
||
print("\n🎤 指令已就绪！支持：")
|
||
print(" - 运动：前进3秒、左转2秒 | - 图像识别：这是什么")
|
||
print(" - 闲聊：今天天气怎么样 | - 音量：增大音量、减小音量\n")
|
||
stream.start()
|
||
current_text[0] = ""
|
||
final_result[0] = ""
|
||
last_command_time[0] = time.time()
|
||
|
||
while True:
|
||
try:
|
||
# 1. 处理音频队列（避免堆积）
|
||
while self.audio_q.qsize() > 5:
|
||
self.audio_q.get_nowait()
|
||
# 2. 发送音频数据（若队列有数据）
|
||
audio_data = self.audio_q.get(timeout=0.5)
|
||
ws.send(audio_data, websocket.ABNF.OPCODE_BINARY)
|
||
|
||
# 3. 指令识别与执行：有文本且2秒内无新音频时执行
|
||
if current_text[0] and (time.time() - last_audio_time[0]) > 2:
|
||
final_result[0] = current_text[0].strip()
|
||
if len(final_result[0]) > 0: # 确保指令有效
|
||
print(f"\n⏹ 最终指令：{final_result[0]}")
|
||
# 调用回调函数执行指令（关键修复：直接在这里执行）
|
||
execute_callback(final_result[0])
|
||
last_command_time[0] = time.time() # 更新最后操作时间
|
||
current_text[0] = "" # 执行后清空，避免重复识别
|
||
final_result[0] = ""
|
||
time.sleep(1) # 等待指令执行完成
|
||
|
||
# 4. 超时检测：30秒无操作则关闭连接
|
||
if time.time() - last_command_time[0] > self.INTERACTION_TIMEOUT:
|
||
print(f"\n⌛ {self.INTERACTION_TIMEOUT}秒无操作，关闭连接")
|
||
self.tts_controller.speak(self.FEEDBACK_TEXT.get("wakeup_timeout", "长时间没操作，我先休息啦"))
|
||
time.sleep(1)
|
||
ws.send("close", websocket.ABNF.OPCODE_TEXT)
|
||
break
|
||
|
||
except queue.Empty:
|
||
# 队列为空时，检测超时
|
||
if time.time() - last_command_time[0] > self.INTERACTION_TIMEOUT:
|
||
print(f"\n⌛ {self.INTERACTION_TIMEOUT}秒无操作，关闭连接")
|
||
self.tts_controller.speak(self.FEEDBACK_TEXT.get("wakeup_timeout", "长时间没操作，我先休息啦"))
|
||
time.sleep(1)
|
||
ws.send("close", websocket.ABNF.OPCODE_TEXT)
|
||
break
|
||
continue # 继续循环等待音频
|
||
except Exception as e:
|
||
print(f"\n❌ 音频发送错误：{str(e)}")
|
||
break
|
||
|
||
audio_thread = threading.Thread(target=send_audio_and_handle, daemon=True)
|
||
audio_thread.start()
|
||
|
||
def start_websocket(self, current_text, final_result, last_audio_time, is_processing, last_command_time, execute_callback):
|
||
"""新增 execute_callback 参数，用于传递指令执行函数"""
|
||
self.stream = sd.RawInputStream(
|
||
samplerate=self.SAMPLE_RATE,
|
||
channels=self.CHANNELS,
|
||
dtype=self.SAMPLE_FORMAT,
|
||
callback=self._audio_callback,
|
||
)
|
||
|
||
ws_url, session_id = self._create_ws_url()
|
||
if not ws_url:
|
||
print("⚠️ 无法生成语音识别连接，3秒后重新监听...")
|
||
time.sleep(3)
|
||
return
|
||
|
||
try:
|
||
print(f"🔄 连接语音识别服务（会话ID：{session_id[:8]}...）")
|
||
# 绑定WebSocket回调时传入 execute_callback
|
||
ws = websocket.WebSocketApp(
|
||
ws_url,
|
||
on_open=lambda ws: self._on_open(ws, self.stream, current_text, final_result, last_audio_time, is_processing, last_command_time, execute_callback),
|
||
on_message=lambda ws, msg: self._on_message(ws, msg, current_text, last_audio_time),
|
||
on_error=lambda ws, err: self._on_error(ws, err, is_processing),
|
||
on_close=lambda ws, status, msg: self._on_close(ws, status, msg, current_text, final_result, self.stream)
|
||
)
|
||
ws.run_forever(ping_interval=10, ping_timeout=5)
|
||
except Exception as e:
|
||
print(f"❌ 语音识别连接失败：{str(e)}")
|
||
print("⚠️ 3秒后重新监听唤醒词...")
|
||
time.sleep(3) |