File size: 5,074 Bytes
c248f01 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 |
# coding=utf-8
# Copyright 2025 The OpenBMB Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import re
import logging
import librosa
import numpy as np
logger = logging.getLogger(__name__)
def is_silent(data):
if np.abs(data).max() < 3e-3:
return True
else:
return False
def sentence_end(txt):
for c in [".", "。", "!", "?", "！", "？"]:
if c in txt:
if c == ".": # check not number before it like 1.
idx = txt.find(c)
if idx > 0:
if txt[idx - 1].isdigit():
continue
return c
return ""
class NumberToTextConverter:
def __init__(self):
self.num_to_chinese = {
"0": "零",
"1": "一",
"2": "二",
"3": "三",
"4": "四",
"5": "五",
"6": "六",
"7": "七",
"8": "八",
"9": "九",
}
self.num_to_english = {
"0": "zero",
"1": "one",
"2": "two",
"3": "three",
"4": "four",
"5": "five",
"6": "six",
"7": "seven",
"8": "eight",
"9": "nine",
}
def number_to_chinese_digit_by_digit(self, num_str):
result = ""
for char in num_str:
if char in self.num_to_chinese:
result += self.num_to_chinese[char]
return result
def number_to_english_digit_by_digit(self, num_str):
result = []
for char in num_str:
if char in self.num_to_english:
result.append(self.num_to_english[char])
return " ".join(result)
def detect_language(self, text):
chinese_count = len(re.findall(r"[\u4e00-\u9fff]", text))
english_count = len(re.findall(r"[a-zA-Z]", text))
return "chinese" if chinese_count >= english_count else "english"
def replace_numbers_with_text(self, text, language=None):
if language is None:
language = self.detect_language(text)
numbers = re.findall(r"\d+", text)
for num in numbers:
if language == "chinese":
replacement = self.number_to_chinese_digit_by_digit(num)
else:
replacement = self.number_to_english_digit_by_digit(num)
text = text.replace(num, replacement, 1)
return text
class VoiceChecker:
def __init__(self):
self.previous_mel = None
self.consecutive_zeros = 0
self.consecutive_low_distance = 0
def compute_distance(self, audio_chunk, mel_spec):
if is_silent(audio_chunk):
return 0.0 # 检查是否为空白片段
mel_db = librosa.power_to_db(mel_spec)
if self.previous_mel is None:
self.previous_mel = mel_db
return -1.0
distance = np.linalg.norm(np.mean(mel_db, axis=1) - np.mean(self.previous_mel, axis=1))
self.previous_mel = mel_db
return distance
def is_bad(self, audio_wav, mel_spec, chunk_size=2560, thresh=100.0):
num_chunks = len(audio_wav) // chunk_size
mel_chunk_size = mel_spec.shape[-1] // num_chunks
for i in range(num_chunks):
audio_chunk = audio_wav[i * chunk_size : (i + 1) * chunk_size]
mel_spec_chunk = mel_spec[:, i * mel_chunk_size : (i + 1) * mel_chunk_size]
distance = self.compute_distance(audio_chunk, mel_spec_chunk)
logger.warning(f"mel dist: {distance:.1f}, zero: {self.consecutive_zeros}, low: {self.consecutive_low_distance}")
if distance == 0:
self.consecutive_low_distance = 0 # reset
self.consecutive_zeros += 1
if self.consecutive_zeros >= 12:
logger.warning("VoiceChecker detected 1.2 s silent. Marking as failed.")
return True
elif distance < thresh:
self.consecutive_zeros = 0
self.consecutive_low_distance += 1
if self.consecutive_low_distance >= 5:
logger.warning("VoiceChecker detected 5 consecutive low distance chunks. Marking as failed.")
return True
else:
self.consecutive_low_distance = 0
self.consecutive_zeros = 0
return False
def reset(self):
self.previous_mel = None
self.consecutive_zeros = 0
self.consecutive_low_distance = 0
|