EngineX-Iluvatar/enginex-mr_series-sherpa-onnx
Archived
16
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

Fix displaying streaming speech recognition results for Python. (#2196)

Browse Source
This commit is contained in:
Fangjun Kuang
2025-05-09 21:48:49 +08:00
committed by GitHub
parent a6834f6556
commit 4a833a7547
8 changed files with 81 additions and 50 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
CMakeLists.txt
View File

@@ -14,7 +14,7 @@ project(sherpa-onnx)
# Remember to update # Remember to update
# ./CHANGELOG.md # ./CHANGELOG.md
# ./new-release.sh # ./new-release.sh
set(SHERPA_ONNX_VERSION "1.11.5") set(SHERPA_ONNX_VERSION "1.11.6")
# Disable warning about # Disable warning about
# #

17
python-api-examples/speech-recognition-from-microphone-with-endpoint-detection-alsa.py
View File

@@ -11,8 +11,8 @@
# to download pre-trained models # to download pre-trained models
import argparse import argparse
import sys
from pathlib import Path from pathlib import Path
import sherpa_onnx import sherpa_onnx
@@ -202,8 +202,8 @@ def main():
stream = recognizer.create_stream() stream = recognizer.create_stream()
last_result = "" display = sherpa_onnx.Display()
segment_id = 0
while True: while True:
samples = alsa.read(samples_per_read) # a blocking read samples = alsa.read(samples_per_read) # a blocking read
stream.accept_waveform(sample_rate, samples) stream.accept_waveform(sample_rate, samples)
@@ -214,13 +214,14 @@ def main():
result = recognizer.get_result(stream) result = recognizer.get_result(stream)
if result and (last_result != result): display.update_text(result)
last_result = result display.display()
print("\r{}:{}".format(segment_id, result), end="", flush=True)
if is_endpoint: if is_endpoint:
if result: if result:
print("\r{}:{}".format(segment_id, result), flush=True) display.finalize_current_sentence()
segment_id += 1 display.display()
recognizer.reset(stream) recognizer.reset(stream)

15
python-api-examples/speech-recognition-from-microphone-with-endpoint-detection.py
View File

@@ -192,8 +192,8 @@ def main():
stream = recognizer.create_stream() stream = recognizer.create_stream()
last_result = "" display = sherpa_onnx.Display()
segment_id = 0
with sd.InputStream(channels=1, dtype="float32", samplerate=sample_rate) as s: with sd.InputStream(channels=1, dtype="float32", samplerate=sample_rate) as s:
while True: while True:
samples, _ = s.read(samples_per_read) # a blocking read samples, _ = s.read(samples_per_read) # a blocking read
@@ -206,13 +206,14 @@ def main():
result = recognizer.get_result(stream) result = recognizer.get_result(stream)
if result and (last_result != result): display.update_text(result)
last_result = result display.display()
print("\r{}:{}".format(segment_id, result), end="", flush=True)
if is_endpoint: if is_endpoint:
if result: if result:
print("\r{}:{}".format(segment_id, result), flush=True) display.finalize_current_sentence()
segment_id += 1 display.display()
recognizer.reset(stream) recognizer.reset(stream)

14
python-api-examples/speech-recognition-from-url.py
View File

@@ -192,8 +192,7 @@ def main():
stream = recognizer.create_stream() stream = recognizer.create_stream()
last_result = "" display = sherpa_onnx.Display()
segment_id = 0
print("Started!") print("Started!")
while True: while True:
@@ -213,13 +212,14 @@ def main():
result = recognizer.get_result(stream) result = recognizer.get_result(stream)
if result and (last_result != result): display.update_text(result)
last_result = result display.display()
print("\r{}:{}".format(segment_id, result), end="", flush=True)
if is_endpoint: if is_endpoint:
if result: if result:
print("\r{}:{}".format(segment_id, result), flush=True) display.finalize_current_sentence()
segment_id += 1 display.display()
recognizer.reset(stream) recognizer.reset(stream)

15
python-api-examples/streaming-paraformer-asr-microphone.py
View File

@@ -74,8 +74,8 @@ def main():
stream = recognizer.create_stream() stream = recognizer.create_stream()
last_result = "" display = sherpa_onnx.Display()
segment_id = 0
with sd.InputStream(channels=1, dtype="float32", samplerate=sample_rate) as s: with sd.InputStream(channels=1, dtype="float32", samplerate=sample_rate) as s:
while True: while True:
samples, _ = s.read(samples_per_read) # a blocking read samples, _ = s.read(samples_per_read) # a blocking read
@@ -88,13 +88,14 @@ def main():
result = recognizer.get_result(stream) result = recognizer.get_result(stream)
if result and (last_result != result): display.update_text(result)
last_result = result display.display()
print("\r{}:{}".format(segment_id, result), end="", flush=True)
if is_endpoint: if is_endpoint:
if result: if result:
print("\r{}:{}".format(segment_id, result), flush=True) display.finalize_current_sentence()
segment_id += 1 display.display()
recognizer.reset(stream) recognizer.reset(stream)

25
python-api-examples/two-pass-speech-recognition-from-microphone.py
View File

@@ -46,7 +46,6 @@ python3 ./python-api-examples/two-pass-speech-recognition-from-microphone.py \
import argparse import argparse
import sys import sys
from pathlib import Path from pathlib import Path
from typing import List
import numpy as np import numpy as np
@@ -375,8 +374,7 @@ def main():
samples_per_read = int(0.1 * sample_rate) # 0.1 second = 100 ms samples_per_read = int(0.1 * sample_rate) # 0.1 second = 100 ms
stream = first_recognizer.create_stream() stream = first_recognizer.create_stream()
last_result = "" display = sherpa_onnx.Display()
segment_id = 0
sample_buffers = [] sample_buffers = []
with sd.InputStream(channels=1, dtype="float32", samplerate=sample_rate) as s: with sd.InputStream(channels=1, dtype="float32", samplerate=sample_rate) as s:
@@ -395,14 +393,8 @@ def main():
result = first_recognizer.get_result(stream) result = first_recognizer.get_result(stream)
result = result.lower().strip() result = result.lower().strip()
if last_result != result: display.update_text(result)
print( display.display()
"\r{}:{}".format(segment_id, " " * len(last_result)),
end="",
flush=True,
)
last_result = result
print("\r{}:{}".format(segment_id, result), end="", flush=True)
if is_endpoint: if is_endpoint:
if result: if result:
@@ -419,14 +411,9 @@ def main():
sample_rate=sample_rate, sample_rate=sample_rate,
) )
result = result.lower().strip() result = result.lower().strip()
display.update_text(result)
print( display.finalize_current_sentence()
"\r{}:{}".format(segment_id, " " * len(last_result)), display.display()
end="",
flush=True,
)
print("\r{}:{}".format(segment_id, result), flush=True)
segment_id += 1
else: else:
sample_buffers = [] sample_buffers = []

2
sherpa-onnx/python/sherpa_onnx/__init__.py
View File

@@ -6,7 +6,6 @@ from _sherpa_onnx import (
AudioTaggingModelConfig, AudioTaggingModelConfig,
CircularBuffer, CircularBuffer,
DenoisedAudio, DenoisedAudio,
Display,
FastClustering, FastClustering,
FastClusteringConfig, FastClusteringConfig,
OfflinePunctuation, OfflinePunctuation,
@@ -48,6 +47,7 @@ from _sherpa_onnx import (
write_wave, write_wave,
) )
from .display import Display
from .keyword_spotter import KeywordSpotter from .keyword_spotter import KeywordSpotter
from .offline_recognizer import OfflineRecognizer from .offline_recognizer import OfflineRecognizer
from .online_recognizer import OnlineRecognizer from .online_recognizer import OnlineRecognizer

41
sherpa-onnx/python/sherpa_onnx/display.py Normal file
View File

@@ -0,0 +1,41 @@
# Copyright (c) 2025 Xiaomi Corporation
import os
from time import gmtime, strftime
def get_current_time():
return strftime("%Y-%m-%d %H:%M:%S", gmtime())
def clear_console():
os.system("cls" if os.name == "nt" else "clear")
class Display:
def __init__(self):
self.sentences = []
self.currentText = ""
def update_text(self, text):
self.currentText = text
def finalize_current_sentence(self):
if self.currentText.strip():
self.sentences.append((get_current_time(), self.currentText))
self.currentText = ""
def display(self):
clear_console()
print("=== Speech Recognition with Next-gen Kaldi ===")
print("Time:", get_current_time())
print("-" * 30)
# display history sentences
if self.sentences:
for i, (when, text) in enumerate(self.sentences):
print(f"[{when}] {i + 1}. {text}")
print("-" * 30)
if self.currentText.strip():
print("Recognizing:", self.currentText)