WIP: saving the optimizing
This commit is contained in:
parent
dbf26dfb36
commit
82f557a4d0
Binary file not shown.
@ -29,7 +29,7 @@
|
||||
"automatic_player_status": true,
|
||||
"automatic_player_url": "http://127.0.0.1:6000",
|
||||
"liveRoom": {
|
||||
"enabled": true,
|
||||
"enabled": false,
|
||||
"url": ""
|
||||
},
|
||||
"record": {
|
||||
|
||||
114
core/recorder.py
114
core/recorder.py
@ -16,8 +16,11 @@ import tempfile
|
||||
import wave
|
||||
from core import fay_core
|
||||
from core import interact
|
||||
|
||||
import re
|
||||
|
||||
# 启动时间 (秒)
|
||||
_ATTACK = 0.2
|
||||
_ATTACK = 0.05 # 更快进入拾音，减少短唤醒词被漏掉的概率
|
||||
|
||||
# 释放时间 (秒)
|
||||
_RELEASE = 0.7
|
||||
@ -85,6 +88,31 @@ class Recorder:
|
||||
with fay_core.auto_play_lock:
|
||||
fay_core.can_auto_play = True
|
||||
|
||||
def _norm_asr_text(self, s: str) -> str:
|
||||
"""
|
||||
大厅场景：ASR 结果常见问题
|
||||
- 句首有空格/标点/语气词（嗯、啊、这个...）
|
||||
- 唤醒词后跟逗号、冒号等
|
||||
所以需要做最小必要的规范化，避免 front 模式误判为“待唤醒”
|
||||
"""
|
||||
if not s:
|
||||
return ""
|
||||
s = s.strip()
|
||||
# 去掉句首标点/空白
|
||||
s = re.sub(r'^[\s，,。.!！？:：、~～]+', '', s)
|
||||
# 去掉句首常见语气词（按需可继续加）
|
||||
s = re.sub(r'^(嗯|啊|呃|额|哦|唔|这个|那个)\s*', '', s)
|
||||
return s
|
||||
|
||||
def _strip_wake_prefix(self, full_text: str, wake_word: str) -> str:
|
||||
"""
|
||||
去除前置唤醒词，只把“真正的问题”交给对话系统
|
||||
例如：'小F，播放音乐' -> '播放音乐'
|
||||
"""
|
||||
rest = full_text[len(wake_word):]
|
||||
# 吞掉唤醒词后的空白/标点
|
||||
return rest.lstrip(" \t，,。.!！？:：、~～")
|
||||
|
||||
def __waitingResult(self, iat: asrclient, audio_data):
|
||||
self.processing = True
|
||||
t = time.time()
|
||||
@ -142,33 +170,82 @@ class Recorder:
|
||||
self.timer = threading.Timer(60, self.reset_wakeup_status) # 重设计时器为60秒
|
||||
self.timer.start()
|
||||
|
||||
#前置唤醒词模式
|
||||
elif cfg.config['source']['wake_word_type'] == 'front':
|
||||
wake_word = cfg.config['source']['wake_word']
|
||||
wake_word_list = wake_word.split(',')
|
||||
# 前置唤醒词模式（大厅优化版）
|
||||
elif cfg.config['source']['wake_word_type'] == 'front':
|
||||
wake_word = cfg.config['source']['wake_word']
|
||||
wake_word_list = [w.strip() for w in wake_word.split(',') if w.strip()]
|
||||
|
||||
raw_text = text
|
||||
text2 = self._norm_asr_text(raw_text)
|
||||
|
||||
wake_up = False
|
||||
for word in wake_word_list:
|
||||
if text.startswith(word):
|
||||
wake_up_word = word
|
||||
matched_word = None
|
||||
|
||||
# 1) 规范化后做“严格句首匹配”
|
||||
for w in wake_word_list:
|
||||
w2 = self._norm_asr_text(w)
|
||||
if w2 and text2.startswith(w2):
|
||||
wake_up = True
|
||||
matched_word = w2
|
||||
break
|
||||
|
||||
# 2) 容错：允许唤醒词出现在句首很短范围内（防止语气词未完全清掉）
|
||||
# 注意：范围要小，避免大厅误唤醒
|
||||
if not wake_up:
|
||||
N = 4 # 建议 3~6，越大越容易误触发
|
||||
head = text2[:N]
|
||||
for w in wake_word_list:
|
||||
w2 = self._norm_asr_text(w)
|
||||
if w2 and w2 in head:
|
||||
wake_up = True
|
||||
matched_word = w2
|
||||
# 从唤醒词出现的位置截断，确保 strip 正确
|
||||
idx = text2.find(w2)
|
||||
text2 = text2[idx:]
|
||||
break
|
||||
|
||||
if wake_up:
|
||||
util.printInfo(1, self.username, "唤醒成功！")
|
||||
util.printInfo(1, self.username, f"唤醒成功！(front:{matched_word})")
|
||||
if wsa_server.get_web_instance().is_connected(self.username):
|
||||
wsa_server.get_web_instance().add_cmd({"panelMsg": "唤醒成功！", "Username" : self.username , 'robot': f'http://{cfg.fay_url}:5000/robot/Listening.jpg'})
|
||||
wsa_server.get_web_instance().add_cmd({
|
||||
"panelMsg": "唤醒成功！",
|
||||
"Username": self.username,
|
||||
'robot': f'http://{cfg.fay_url}:5000/robot/Listening.jpg'
|
||||
})
|
||||
if wsa_server.get_instance().is_connected(self.username):
|
||||
content = {'Topic': 'Unreal', 'Data': {'Key': 'log', 'Value': "唤醒成功！"}, 'Username' : self.username, 'robot': f'http://{cfg.fay_url}:5000/robot/Listening.jpg'}
|
||||
content = {
|
||||
'Topic': 'Unreal',
|
||||
'Data': {'Key': 'log', 'Value': "唤醒成功！"},
|
||||
'Username': self.username,
|
||||
'robot': f'http://{cfg.fay_url}:5000/robot/Listening.jpg'
|
||||
}
|
||||
wsa_server.get_instance().add_cmd(content)
|
||||
#去除唤醒词后语句
|
||||
question = text#[len(wake_up_word):].lstrip()
|
||||
|
||||
# ✅ 关键：剥离唤醒词，把真正问题交给对话系统
|
||||
question = self._strip_wake_prefix(text2, matched_word)
|
||||
|
||||
# 如果只说了唤醒词（或后面太短），给一句提示
|
||||
if not question:
|
||||
question = "在呢，你说？"
|
||||
|
||||
self.on_speaking(question)
|
||||
self.processing = False
|
||||
else:
|
||||
util.printInfo(1, self.username, "[!] 待唤醒！")
|
||||
# ✅ 关键：打印原始识别和规范化后文本，现场好定位为何没匹配上
|
||||
util.printInfo(1, self.username, f"[!] 待唤醒！(front) ASR='{raw_text}' norm='{text2}'")
|
||||
if wsa_server.get_web_instance().is_connected(self.username):
|
||||
wsa_server.get_web_instance().add_cmd({"panelMsg": "[!] 待唤醒！", "Username" : self.username , 'robot': f'http://{cfg.fay_url}:5000/robot/Normal.jpg'})
|
||||
wsa_server.get_web_instance().add_cmd({
|
||||
"panelMsg": "[!] 待唤醒！",
|
||||
"Username": self.username,
|
||||
'robot': f'http://{cfg.fay_url}:5000/robot/Normal.jpg'
|
||||
})
|
||||
if wsa_server.get_instance().is_connected(self.username):
|
||||
content = {'Topic': 'Unreal', 'Data': {'Key': 'log', 'Value': "[!] 待唤醒！"}, 'Username' : self.username, 'robot': f'http://{cfg.fay_url}:5000/robot/Normal.jpg'}
|
||||
content = {
|
||||
'Topic': 'Unreal',
|
||||
'Data': {'Key': 'log', 'Value': "[!] 待唤醒！"},
|
||||
'Username': self.username,
|
||||
'robot': f'http://{cfg.fay_url}:5000/robot/Normal.jpg'
|
||||
}
|
||||
wsa_server.get_instance().add_cmd(content)
|
||||
|
||||
#非唤醒模式
|
||||
@ -234,7 +311,10 @@ class Recorder:
|
||||
|
||||
#计算音量是否满足激活拾音
|
||||
level = audioop.rms(data, 2)
|
||||
if len(self.__history_data) >= 10:#保存激活前的音频，以免信息掉失
|
||||
|
||||
# 把激活前缓存拉长，避免“唤醒词”在触发拾音前被漏掉
|
||||
# 1024帧@16kHz≈64ms/块，30块≈1.9秒
|
||||
if len(self.__history_data) >= 30: # 保存激活前的音频，以免信息掉失
|
||||
self.__history_data.pop(0)
|
||||
if len(self.__history_level) >= 500:
|
||||
self.__history_level.pop(0)
|
||||
|
||||
@ -150,6 +150,19 @@ class FayInterface {
|
||||
|
||||
handleIncomingMessage(data) {
|
||||
const vueInstance = this.vueInstance;
|
||||
|
||||
if (data.panelReply !== undefined) {
|
||||
vueInstance.panelReply = data.panelReply.content;
|
||||
|
||||
// 发送消息给父窗口，并指定目标 origin（必须是父组件的域名）
|
||||
if (window.parent) {
|
||||
window.parent.postMessage(
|
||||
{type: 'panelReply', data: data.panelReply.content},
|
||||
'*' // 父组件的域名
|
||||
);
|
||||
}
|
||||
}
|
||||
|
||||
// console.log('Incoming message:', data);
|
||||
if (data.liveState !== undefined) {
|
||||
vueInstance.liveState = data.liveState;
|
||||
|
||||
2
qa.csv
2
qa.csv
@ -1,4 +1,4 @@
|
||||
你好,你好,我是小橄榄！有什么我可以帮助你的吗
|
||||
你好,你好，我是小橄榄！有什么我可以帮助你的吗
|
||||
我们现在在哪里,我们现在在冕宁元升农业的展览厅，可以参观游览我们先进的油橄榄产业园区哦！
|
||||
介绍一下基地,元升集团油橄榄种植基地是中国目前最大的油橄榄种植庄园，目前整个庄园面积已接近30000亩。
|
||||
介绍一下元升集团,2011年，冕宁元升农业董事长林春福跟随周恩来总理的脚步，在冕宁地区开启了油橄榄庄园打造之路。经过十年的发展，冕宁元升农业目前已成为国家林业局示范基地、国家林业重点龙头企业、四川省第一种植庄园、四川省脱贫标杆企业，获得各种荣誉奖项200余项。
|
||||
|
||||
|
@ -52,11 +52,11 @@ lingju_api_authcode=
|
||||
|
||||
#gpt 服务密钥(NLP多选1) https://openai.com/
|
||||
#免费key只支持gpt 3.5 ，若想使用其他model，可到 https://api.zyai.online/register/?aff_code=MyCI 下购买申请。
|
||||
gpt_api_key=sk-4Spva89SGSikpacz3a70Dd081cA84c9a8dEd345f19C9BdFc
|
||||
gpt_api_key=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
|
||||
#gpt base url 如：https://api.openai.com/v1、https://rwkv.ai-creator.net/chntuned/v1、https://api.fastgpt.in/api/v1、https://api.moonshot.cn/v1
|
||||
gpt_base_url=https://api.zyai.online/v1
|
||||
gpt_base_url=https://openrouter.ai/api/v1
|
||||
#gpt model engine 如：gpt-3.5-turbo、moonshot-v1-8k
|
||||
gpt_model_engine=gpt-3.5-turbo
|
||||
gpt_model_engine=qwen/qwen3-4b:free
|
||||
|
||||
#gpt(fastgpt)代理(可为空，填写例子：127.0.0.1:7890)
|
||||
proxy_config=
|
||||
|
||||
Loading…
Reference in New Issue
Block a user