import streamlit as st
import speech_recognition as sr
import librosa
import numpy as np
import noisereduce as nr
import soundfile as sf
import tempfile
import opensmile

# Custom CSS for a modern UI
st.markdown("""
    <style>
        body {
            background-color: #f5f5f5;
        }
        .stApp {
            background-color: #ffffff;
            border-radius: 10px;
            padding: 20px;
            box-shadow: 0px 0px 10px rgba(0, 0, 0, 0.1);
        }
        h1 {
            color: #FF5733;
            text-align: center;
        }
        .button {
            background-color: #ff6600;
            color: white;
            border-radius: 5px;
            padding: 10px;
            font-size: 16px;
            font-weight: bold;
        }
        .button:hover {
            background-color: #cc5200;
        }
    </style>
""", unsafe_allow_html=True)

# Initialize OpenSMILE for emotion detection
smile = opensmile.Smile(
    feature_set=opensmile.FeatureSet.ComParE_2016,
    feature_level=opensmile.FeatureLevel.Functionals,
)

# Streamlit UI
st.markdown("<h1>🎙️ Speech Detection System</h1>", unsafe_allow_html=True)
st.write("🔹 Record speech and analyze it in real time.")

# Audio Recorder
st.markdown("## 🎤 Record Your Voice")
recognizer = sr.Recognizer()

with sr.Microphone() as source:
    st.write("Press **Start Recording** and speak...")
    if st.button("🎙️ Start Recording", key="record", help="Click to start recording"):
        with st.spinner("Listening..."):
            audio = recognizer.listen(source)
            st.success("🎧 Recording Complete!")
            with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_audio:
                temp_audio.write(audio.get_wav_data())
                audio_path = temp_audio.name

        # Speech Recognition
        st.markdown("## 📜 Speech-to-Text")
        try:
            text = recognizer.recognize_google(audio)
            st.success(f"🗣️ Recognized Speech: **{text}**")
        except sr.UnknownValueError:
            st.warning("⚠️ Speech not recognized. Please try again.")

        # Acoustic Analysis
        st.markdown("## 📊 Acoustic Analysis")
        y, sr_val = librosa.load(audio_path)

        # Pitch Calculation
        pitch, _ = librosa.piptrack(y=y, sr=sr_val)
        mean_pitch = np.mean(pitch[pitch > 0])
        st.write(f"🎵 **Mean Pitch:** {mean_pitch:.2f} Hz")

        # Volume Calculation
        rms = librosa.feature.rms(y=y)
        mean_volume = np.mean(rms)
        st.write(f"🔊 **Mean Volume:** {mean_volume:.2f}")

        # Noise Filtering
        st.markdown("## 🔇 Noise Reduction")
        reduced_noise = nr.reduce_noise(y=y, sr=sr_val)
        noise_filtered_path = "cleaned_audio.wav"
        sf.write(noise_filtered_path, reduced_noise, sr_val)
        st.audio(noise_filtered_path, format="audio/wav")
        st.success("✅ Noise Reduced Successfully!")

        # Emotional Detection
        st.markdown("## 😊 Emotion Detection")
        features = smile.process_file(audio_path)
        st.write("🔍 **Emotion Features Extracted!**")
        st.dataframe(features)

        st.success("✅ Speech Analysis Completed!")