File size: 4,395 Bytes
9e34ea2 | 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 | from flask import Flask, request, jsonify
from flask_cors import CORS
from werkzeug.utils import secure_filename
import os
import time
import uuid
import librosa
import numpy as np
import joblib
app = Flask(__name__)
# ── CORS ──────────────────────────────────────────────────────────
# Restrict this to the actual domain(s) that serve your frontend once
# you know them (e.g. your Hugging Face Space / GitHub Pages URL).
# Using "*" (allow-all) is fine for local testing only.
ALLOWED_ORIGINS = os.environ.get("ALLOWED_ORIGINS", "*")
CORS(app, origins=ALLOWED_ORIGINS.split(",") if ALLOWED_ORIGINS != "*" else "*")
# ── Upload settings ───────────────────────────────────────────────
UPLOAD_FOLDER = "uploads"
os.makedirs(UPLOAD_FOLDER, exist_ok=True)
ALLOWED_EXTENSIONS = {"wav", "mp3"}
MAX_CONTENT_LENGTH = 10 * 1024 * 1024 # 10 MB per request, adjust as needed
app.config["MAX_CONTENT_LENGTH"] = MAX_CONTENT_LENGTH
def allowed_file(filename):
return "." in filename and filename.rsplit(".", 1)[1].lower() in ALLOWED_EXTENSIONS
# Load trained model once
model = joblib.load("stutter_model.pkl")
def extract_features(file_path):
"""Same feature extraction used during training: 40 MFCCs averaged over time."""
try:
audio, sr = librosa.load(file_path, duration=5)
if len(audio) == 0:
return None
mfcc = librosa.feature.mfcc(
y=audio,
sr=sr,
n_mfcc=40
)
features = np.mean(mfcc.T, axis=0)
return features
except Exception as e:
print("Feature Error:", e)
return None
@app.route("/")
def home():
return "Speech Detection AI Backend Running ✅"
@app.route("/api/predict", methods=["POST"])
def predict():
filepath = None
try:
file = request.files.get("file")
if not file or file.filename == "":
return jsonify({"error": "No file uploaded"}), 400
# Validate extension before touching the filesystem
if not allowed_file(file.filename):
return jsonify({"error": "Unsupported file type. Only .wav and .mp3 are allowed."}), 400
# Sanitize the filename and make it unique to prevent path
# traversal and filename collisions between concurrent users.
safe_name = secure_filename(file.filename)
ext = safe_name.rsplit(".", 1)[1].lower()
unique_name = f"{uuid.uuid4().hex}.{ext}"
filepath = os.path.join(UPLOAD_FOLDER, unique_name)
file.save(filepath)
start_time = time.time()
# Extract MFCC features
features = extract_features(filepath)
if features is None:
return jsonify({"error": "Feature extraction failed"}), 500
# Model Prediction
prediction = model.predict([features])[0]
probabilities = model.predict_proba([features])[0]
processing_ms = int((time.time() - start_time) * 1000)
normal_prob = round(probabilities[0] * 100, 2)
stutter_prob = round(probabilities[1] * 100, 2)
confidence = round(max(normal_prob, stutter_prob), 2)
result = {
"isNormal": bool(prediction == 0),
"confidence": confidence,
"fluency": confidence,
"processingMs": processing_ms,
"normalProb": normal_prob,
"stutterProb": stutter_prob
}
return jsonify(result)
except Exception as e:
# Avoid leaking internal details (stack traces, paths) to the client
print("Prediction error:", e)
return jsonify({"error": "Something went wrong while processing the file."}), 500
finally:
# clean up uploaded temp file
try:
if filepath and os.path.exists(filepath):
os.remove(filepath)
except Exception:
pass
if __name__ == "__main__":
# debug=False is required for any public/shared deployment — Flask's
# debugger allows remote code execution if it's ever reachable.
debug_mode = os.environ.get("FLASK_DEBUG", "false").lower() == "true"
app.run(debug=debug_mode, host="0.0.0.0", port=int(os.environ.get("PORT", 5000)))
|