public/Telereview
1
0
Fork 0
mirror of https://git.roussel.pro/telecom-paris/pact.git synced 2026-02-09 02:20:17 +01:00
Code Issues Actions Packages Projects Releases Wiki Activity

intégration de la détéction audio à l'interface

Browse Source
This commit is contained in:
Quentin Roussel
2023-03-26 23:48:17 +02:00
parent f978ed0a8b
commit 38c9e4e0ea
23 changed files with 74 additions and 40 deletions
Download Patch File Download Diff File Expand all files Collapse all files

19
code/backend_reconnaissance/audio_detector.py
View File

@@ -119,7 +119,6 @@ def init_database():
if not os.path.isfile(os.path.join(data_dir, word)):
for file in os.listdir(os.path.join(data_dir,word)):
if os.path.isfile(os.path.join(data_dir, word,file)):
print(word,os.path.join(data_dir, word,file))
words.append(word)
files.append(os.path.join(data_dir, word,file))
return words,files
@@ -130,7 +129,23 @@ def get_word_metadata(word):
return data[word]
#Todo : detecte si pas de note donnée
def get_grade():
def record():
sr = 44100 # fréquence d'échantillonnage
duration = 6 # durée d'enregistrement en secondes
filename = "recording" # nom du fichier à enregistrer
record_audio(filename, duration, sr)
audio_query,sr = librosa.load(f'{filename}.wav', sr=sr)
return audio_query,sr
def analyze(audio_query,sr):
coupe_silence(audio_query)
words, files = init_database()
audio_train_list = [librosa.load(file, sr=sr)[0] for file in files]
recognized_word_index = recognize_speech(audio_query, audio_train_list, sr)
recognized_word = words[recognized_word_index]
return get_word_metadata(recognized_word)
def test():
sr = 44100 # fréquence d'échantillonnage
duration = 6 # durée d'enregistrement en secondes
filename = "recording" # nom du fichier à enregistrer