python实时语音识别 -2025.12.29

基于科大讯飞的语音识别大模型

test代码

import base64
import hashlib
import hmac
import time
import uuid
import urllib.parse
import json
import threading

import pyaudio
from websocket import WebSocketApp

APP_ID = "xxx"
API_SECRET = "xxx"
API_KEY = "xxx"

WS_HOST = "wss://office-api-ast-dx.iflyaisol.com/ast/communicate/v1"
LANG = "autodialect"
AUDIO_ENCODE = "pcm_s16le"
SAMPLE_RATE = 16000
UUID = uuid.uuid4().hex

FRAME_BYTES = 1280
FRAME_INTERVAL = 0.04
session_id = str(uuid.uuid4())

# ---------------------
# 签名生成
# ---------------------
def build_signature(app_id, api_key, api_secret, lang, audio_encode, samplerate, uuid):
    utc_raw = time.strftime("%Y-%m-%dT%H:%M:%S+0800", time.localtime())
    params = {
        "accessKeyId": api_key,
        "appId": app_id,
        "lang": lang,
        "utc": utc_raw,
        "uuid": uuid,
        "audio_encode": audio_encode,
        "samplerate": str(samplerate),
    }
    sorted_items = sorted(params.items(), key=lambda x: x[0])
    encoded_pairs = []
    for k, v in sorted_items:
        k_enc = urllib.parse.quote(k, safe="")
        v_enc = urllib.parse.quote(v, safe="")
        encoded_pairs.append(f"{k_enc}={v_enc}")
    base_string = "&".join(encoded_pairs)
    digest = hmac.new(api_secret.encode("utf-8"),
                      base_string.encode("utf-8"),
                      hashlib.sha1).digest()
    signature = base64.b64encode(digest).decode("utf-8")
    return params, signature

def build_ws_url():
    params, signature = build_signature(
        APP_ID, API_KEY, API_SECRET, LANG, AUDIO_ENCODE, SAMPLE_RATE, UUID
    )
    query_items = []
    for k, v in params.items():
        query_items.append(f"{urllib.parse.quote(k, safe='')}={urllib.parse.quote(v, safe='')}")
    query_items.append(f"signature={urllib.parse.quote(signature, safe='')}")
    query_string = "&".join(query_items)
    return f"{WS_HOST}?{query_string}"

# ---------------------
# WebSocket 事件处理
# ---------------------
final_results = {}  # 缓冲区：seg_id -> text

def on_open(ws):
    print("[WS] 连接已打开，开始麦克风采集...")
    def run():
        pa = pyaudio.PyAudio()
        stream = pa.open(format=pyaudio.paInt16,
                         channels=1,
                         rate=SAMPLE_RATE,
                         input=True,
                         frames_per_buffer=FRAME_BYTES)
        try:
            while True:
                data = stream.read(FRAME_BYTES, exception_on_overflow=False)
                if not data:
                    break
                ws.send(data, opcode=0x2)
                time.sleep(FRAME_INTERVAL)
        except KeyboardInterrupt:
            print("[WS] 采集中断")
        except Exception as e:
            print(f"[WS] 采集异常: {e}")
        finally:
            stream.stop_stream()
            stream.close()
            pa.terminate()
            end_msg = {"end": True, "sessionId": session_id}
            ws.send(json.dumps(end_msg))
            print("[WS] 已发送结束标识")

    threading.Thread(target=run, daemon=True).start()

def on_message(ws, message):
    try:
        obj = json.loads(message)
    except Exception:
        print("[WS] 非JSON消息:", message[:200])
        return

    msg_type = obj.get("msg_type")
    res_type = obj.get("res_type")
    data = obj.get("data", {})

    if msg_type == "result" and res_type == "asr":
        seg_id = data.get("seg_id")
        cn = data.get("cn", {})
        st = cn.get("st", {})
        result_type = st.get("type")  # 0=最终结果, 1=中间结果
        rt_list = st.get("rt", [])
        text_fragments = []
        for rt in rt_list:
            for ws_item in rt.get("ws", []):
                for cw in ws_item.get("cw", []):
                    w = cw.get("w")
                    if w:
                        text_fragments.append(w)
        text = "".join(text_fragments)

        if result_type == "0":  # 只处理最终结果
            final_results[seg_id] = text
            # 按 seg_id 顺序拼接完整文本
            ordered_text = "".join(final_results[i] for i in sorted(final_results.keys()))
            print(f"[FINAL] {ordered_text}")

    elif msg_type == "result" and res_type == "frc":
        desc = data.get("desc")
        detail = data.get("detail")
        print(f"[ERR] 异常: {desc}, detail={detail}")
    else:
        print("[WS] 其他消息:", obj)

def on_error(ws, error):
    print("[WS] 连接错误:", error)

def on_close(ws, close_status_code, close_msg):
    print(f"[WS] 连接关闭: code={close_status_code}, msg={close_msg}")

def main():
    ws_url = build_ws_url()
    print("WebSocket URL:", ws_url)
    ws_app = WebSocketApp(
        ws_url,
        on_open=on_open,
        on_message=on_message,
        on_error=on_error,
        on_close=on_close
    )
    ws_app.run_forever()

if __name__ == "__main__":
    main()