From 4870f0f8a5f3ae9072a56b6f320ba7ebcfaf4373 Mon Sep 17 00:00:00 2001
From: Binbin Gu <gubinbin@outlook.com>
Date: 星期五, 02 六月 2023 11:51:02 +0800
Subject: [PATCH] Update cardinal.py (#562)
---
funasr/models/e2e_vad.py | 14 +++++---------
1 files changed, 5 insertions(+), 9 deletions(-)
diff --git a/funasr/models/e2e_vad.py b/funasr/models/e2e_vad.py
index e477750..71ed2cf 100644
--- a/funasr/models/e2e_vad.py
+++ b/funasr/models/e2e_vad.py
@@ -40,7 +40,6 @@
Deep-FSMN for Large Vocabulary Continuous Speech Recognition
https://arxiv.org/abs/1803.05030
"""
-
def __init__(
self,
sample_rate: int = 16000,
@@ -110,7 +109,6 @@
Deep-FSMN for Large Vocabulary Continuous Speech Recognition
https://arxiv.org/abs/1803.05030
"""
-
def __init__(self):
self.start_ms = 0
self.end_ms = 0
@@ -134,7 +132,6 @@
Deep-FSMN for Large Vocabulary Continuous Speech Recognition
https://arxiv.org/abs/1803.05030
"""
-
def __init__(self):
self.noise_prob = 0.0
self.speech_prob = 0.0
@@ -149,7 +146,6 @@
Deep-FSMN for Large Vocabulary Continuous Speech Recognition
https://arxiv.org/abs/1803.05030
"""
-
def __init__(self, window_size_ms: int, sil_to_speech_time: int,
speech_to_sil_time: int, frame_size_ms: int):
self.window_size_ms = window_size_ms
@@ -221,7 +217,6 @@
Deep-FSMN for Large Vocabulary Continuous Speech Recognition
https://arxiv.org/abs/1803.05030
"""
-
def __init__(self, encoder: FSMN, vad_post_args: Dict[str, Any], frontend=None):
super(E2EVadModel, self).__init__()
self.vad_opts = VADXOptions(**vad_post_args)
@@ -231,7 +226,6 @@
self.vad_opts.frame_in_ms)
self.encoder = encoder
# init variables
- self.is_final = False
self.data_buf_start_frame = 0
self.frm_cnt = 0
self.latest_confirmed_speech_frame = 0
@@ -262,7 +256,6 @@
self.frontend = frontend
def AllResetDetection(self):
- self.is_final = False
self.data_buf_start_frame = 0
self.frm_cnt = 0
self.latest_confirmed_speech_frame = 0
@@ -478,6 +471,8 @@
def forward(self, feats: torch.Tensor, waveform: torch.tensor, in_cache: Dict[str, torch.Tensor] = dict(),
is_final: bool = False
) -> Tuple[List[List[List[int]]], Dict[str, torch.Tensor]]:
+ if not in_cache:
+ self.AllResetDetection()
self.waveform = waveform # compute decibel for each frame
self.ComputeDecibel()
self.ComputeScores(feats, in_cache)
@@ -490,8 +485,7 @@
segment_batch = []
if len(self.output_data_buf) > 0:
for i in range(self.output_data_buf_offset, len(self.output_data_buf)):
- if not is_final and (
- not self.output_data_buf[i].contain_seg_start_point or not self.output_data_buf[
+ if not is_final and (not self.output_data_buf[i].contain_seg_start_point or not self.output_data_buf[
i].contain_seg_end_point):
continue
segment = [self.output_data_buf[i].start_ms, self.output_data_buf[i].end_ms]
@@ -507,6 +501,8 @@
def forward_online(self, feats: torch.Tensor, waveform: torch.tensor, in_cache: Dict[str, torch.Tensor] = dict(),
is_final: bool = False, max_end_sil: int = 800
) -> Tuple[List[List[List[int]]], Dict[str, torch.Tensor]]:
+ if not in_cache:
+ self.AllResetDetection()
self.max_end_sil_frame_cnt_thresh = max_end_sil - self.vad_opts.speech_to_sil_time_thres
self.waveform = waveform # compute decibel for each frame
--
Gitblit v1.9.1