From a917d7557dd2b1e5263eeba7e5e4d5a5fc02f69f Mon Sep 17 00:00:00 2001
From: 游雁 <zhifu.gzf@alibaba-inc.com>
Date: 星期四, 27 四月 2023 11:41:16 +0800
Subject: [PATCH] websocket
---
funasr/runtime/python/websocket/ASR_client.py | 71 +++++++++++++++++++++++++++++------
1 files changed, 58 insertions(+), 13 deletions(-)
diff --git a/funasr/runtime/python/websocket/ASR_client.py b/funasr/runtime/python/websocket/ASR_client.py
index fa95328..9a4a148 100644
--- a/funasr/runtime/python/websocket/ASR_client.py
+++ b/funasr/runtime/python/websocket/ASR_client.py
@@ -1,9 +1,7 @@
-import pyaudio
-# import websocket #鍖哄埆鏈嶅姟绔繖閲屾槸 websocket-client搴�
+# -*- encoding: utf-8 -*-
import time
import websockets
import asyncio
-from queue import Queue
# import threading
import argparse
import json
@@ -30,12 +28,13 @@
args = parser.parse_args()
+# voices = asyncio.Queue()
+from queue import Queue
voices = Queue()
-
-
# 鍏朵粬鍑芥暟鍙互閫氳繃璋冪敤send(data)鏉ュ彂閫佹暟鎹紝渚嬪锛�
-async def record():
+async def record_microphone():
+ import pyaudio
#print("2")
global voices
FORMAT = pyaudio.paInt16
@@ -50,17 +49,60 @@
rate=RATE,
input=True,
frames_per_buffer=CHUNK)
-
+ is_speaking = True
while True:
data = stream.read(CHUNK)
+ data = data.decode('ISO-8859-1')
+ message = json.dumps({"chunk": args.chunk_size, "is_speaking": is_speaking, "audio": data})
- voices.put(data)
+ voices.put(message)
#print(voices.qsize())
- await asyncio.sleep(0.01)
-
+ await asyncio.sleep(0.005)
+# 鍏朵粬鍑芥暟鍙互閫氳繃璋冪敤send(data)鏉ュ彂閫佹暟鎹紝渚嬪锛�
+async def record_from_scp():
+ import wave
+ global voices
+ if args.audio_in.endswith(".scp"):
+ f_scp = open(args.audio_in)
+ wavs = f_scp.readlines()
+ else:
+ wavs = [args.audio_in]
+ for wav in wavs:
+ wav_splits = wav.strip().split()
+ wav_path = wav_splits[1] if len(wav_splits) > 1 else wav_splits[0]
+ # bytes_f = open(wav_path, "rb")
+ # bytes_data = bytes_f.read()
+ with wave.open(wav_path, "rb") as wav_file:
+ # 鑾峰彇闊抽鍙傛暟
+ params = wav_file.getparams()
+ # 鑾峰彇澶翠俊鎭殑闀垮害
+ # header_length = wav_file.getheaders()[0][1]
+ # 璇诲彇闊抽甯ф暟鎹紝璺宠繃澶翠俊鎭�
+ # wav_file.setpos(header_length)
+ frames = wav_file.readframes(wav_file.getnframes())
+
+ # 灏嗛煶棰戝抚鏁版嵁杞崲涓哄瓧鑺傜被鍨嬬殑鏁版嵁
+ audio_bytes = bytes(frames)
+ stride = int(args.chunk_size/1000*16000*2)
+ chunk_num = (len(audio_bytes)-1)//stride + 1
+ print(stride)
+ is_speaking = True
+ for i in range(chunk_num):
+ if i == chunk_num-1:
+ is_speaking = False
+ beg = i*stride
+ data = audio_bytes[beg:beg+stride]
+ data = data.decode('ISO-8859-1')
+ message = json.dumps({"chunk": args.chunk_size, "is_speaking": is_speaking, "audio": data})
+ voices.put(message)
+ # print("data_chunk: ", len(data_chunk))
+ # print(voices.qsize())
+
+ await asyncio.sleep(args.chunk_size/1000)
+
async def ws_send():
global voices
@@ -74,8 +116,8 @@
await websocket.send(data) # 閫氳繃ws瀵硅薄鍙戦�佹暟鎹�
except Exception as e:
print('Exception occurred:', e)
- await asyncio.sleep(0.01)
- await asyncio.sleep(0.01)
+ await asyncio.sleep(0.005)
+ await asyncio.sleep(0.005)
@@ -97,7 +139,10 @@
uri = "ws://{}:{}".format(args.host, args.port)
#ws = await websockets.connect(uri, subprotocols=["binary"]) # 鍒涘缓涓�涓暱杩炴帴
async for websocket in websockets.connect(uri, subprotocols=["binary"], ping_interval=None):
- task = asyncio.create_task(record()) # 鍒涘缓涓�涓悗鍙颁换鍔″綍闊�
+ if args.audio_in is not None:
+ task = asyncio.create_task(record_from_scp()) # 鍒涘缓涓�涓悗鍙颁换鍔″綍闊�
+ else:
+ task = asyncio.create_task(record_microphone()) # 鍒涘缓涓�涓悗鍙颁换鍔″綍闊�
task2 = asyncio.create_task(ws_send()) # 鍒涘缓涓�涓悗鍙颁换鍔″彂閫�
task3 = asyncio.create_task(message()) # 鍒涘缓涓�涓悗鍙版帴鏀舵秷鎭殑浠诲姟
await asyncio.gather(task, task2, task3)
--
Gitblit v1.9.1