ser_detection.py

from __future__ import absolute_import, division, print_function, unicode_literals

from flask import Flask, make_response, render_template, request, jsonify, redirect, url_for, send_from_directory
from flask_cors import CORS

import sys
import os

import librosa
import librosa.display
import numpy as np

import warnings
import tensorflow as tf
from keras.models import Sequential
from keras.layers import Dense
from keras.utils import to_categorical
from keras.layers import Flatten, Dropout, Activation
from keras.layers import Conv2D, MaxPooling2D
from keras.layers.normalization import BatchNormalization
from sklearn.model_selection import train_test_split
from tqdm import tqdm
# import scipy.io.wavfile as wav
# from speechpy.feature import mfcc

import pyaudio
import wave

warnings.filterwarnings("ignore")

app = Flask(__name__)
CORS(app)

classLabels = ('Angry', 'Fear', 'Disgust', 'Happy', 'Sad', 'Surprised', 'Neutral')
numLabels = len(classLabels)
in_shape = (39,216)
model = Sequential()

model.add(Conv2D(8, (13, 13), input_shape=(in_shape[0], in_shape[1], 1)))
model.add(BatchNormalization(axis=-1))
model.add(Activation('relu'))
model.add(Conv2D(8, (13, 13)))
model.add(BatchNormalization(axis=-1))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 1)))
model.add(Conv2D(8, (3, 3)))
model.add(BatchNormalization(axis=-1))
model.add(Activation('relu'))
model.add(Conv2D(8, (1, 1)))
model.add(BatchNormalization(axis=-1))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 1)))
model.add(Flatten())
model.add(Dense(64))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(Dropout(0.2))

model.add(Dense(numLabels, activation='softmax'))
model.compile(loss='binary_crossentropy', optimizer='adam',
                           metrics=['accuracy'])
# print(model.summary(), file=sys.stderr)

model.load_weights('speech_emotion_detection_ravdess_savee.h5')

def detect_emotion(file_name):
    X, sample_rate = librosa.load(file_name, res_type='kaiser_best',duration=2.5,sr=22050*2,offset=0.5)
    sample_rate = np.array(sample_rate)
    mfccs = librosa.feature.mfcc(y=X, sr=sample_rate, n_mfcc=39)
    feature = mfccs
    print("Feature_shape =>",feature.shape)
    feature = feature.reshape(39, 216, 1)
    result = classLabels[np.argmax(model.predict(np.array([feature])))]
    print("Result ==> ",result) 
    return result

@app.route("/speech-emotion-recognition/")
def emotion_detection():
    filename = 'audio_files/Happy.wav'

    result = detect_emotion(filename)
    return result

@app.route("/record_audio/")
def record_audio():
    CHUNK = 1024 
    FORMAT = pyaudio.paInt16 #paInt8
    CHANNELS = 2 
    RATE = 44100 #sample rate
    RECORD_SECONDS = 4
    
    fileList = os.listdir('recorded_audio') 
    print("Audio File List ==> ",fileList)

    new_wav_file = ""
    
    if(fileList):
        filename_list = []
        for i in fileList:
            print(i)
            filename = i.split('.')[0]
            filename_list.append(filename)
        
        max_file = max(filename_list)
        print(type(max_file))

        new_wav_file = int(max_file) + 1
    else:
        new_wav_file="1"

    new_wav_file = str(new_wav_file) + ".wav"
    filepath = os.path.join('recorded_audio', new_wav_file)
    WAVE_OUTPUT_FILENAME = filepath

    print(WAVE_OUTPUT_FILENAME)

    p = pyaudio.PyAudio()

    stream = p.open(format=FORMAT,
                    channels=CHANNELS,
                    rate=RATE,
                    input=True,
                    frames_per_buffer=CHUNK) #buffer

    print("* recording")

    frames = []

    for i in range(0, int(RATE / CHUNK * RECORD_SECONDS)):
        data = stream.read(CHUNK)
        frames.append(data) # 2 bytes(16 bits) per channel

    print("* done recording")

    stream.stop_stream()
    stream.close()
    p.terminate()

    wf = wave.open(WAVE_OUTPUT_FILENAME, 'wb')
    wf.setnchannels(CHANNELS)
    wf.setsampwidth(p.get_sample_size(FORMAT))
    wf.setframerate(RATE)
    wf.writeframes(b''.join(frames))
    wf.close()
    return "Audio Recorded"

if __name__ == "__main__":
    app.run()