app.py

from fastapi import FastAPI, File, UploadFile
import librosa
import numpy as np
import shutil
from funasr import AutoModel
from funasr.utils.postprocess_utils import rich_transcription_postprocess

app = FastAPI()

# Load mô hình SenseVoiceSmall
model_dir = "FunAudioLLM/SenseVoiceSmall"


model = AutoModel(
    model=model_dir,
    vad_model="fsmn-vad",
    vad_kwargs={"max_single_segment_time": 30000},
    device="cuda:0",
    hub="hf",
)

# Hàm tính RMS energy
def calculate_rms_energy(audio_path):
    y, sr = librosa.load(audio_path)
    rms = librosa.feature.rms(y=y)[0]
    return np.mean(rms)

# Hàm phát hiện tiếng ồn
def detect_noise(audio_path):
    rms_energy = calculate_rms_energy(audio_path)
    res = model.generate(input=audio_path, language="auto", audio_event_detection=True)
    audio_events = res[0].get("audio_event_detection", {})

    if rms_energy > 0.02:
        return "ồn ào"
    elif rms_energy > 0.01:
        for event_label, event_score in audio_events.items():
            if event_score > 0.7 and event_label in ["laughter", "applause", "crying", "coughing"]:
                return f"ồn ào ({event_label})"
    return "yên tĩnh"

# API nhận file âm thanh từ Flutter
@app.post("/detect-noise/")
async def detect_noise_api(file: UploadFile = File(...)):
    file_path = f"temp/{file.filename}"
    with open(file_path, "wb") as buffer:
        shutil.copyfileobj(file.file, buffer)

    result = detect_noise(file_path)
    return {"noise_level": result}