{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"## Import the libraries\n","metadata":{"execution":{"iopub.status.busy":"2022-02-11T15:15:42.09471Z","iopub.execute_input":"2022-02-11T15:15:42.095103Z","iopub.status.idle":"2022-02-11T15:15:42.099724Z","shell.execute_reply.started":"2022-02-11T15:15:42.095016Z","shell.execute_reply":"2022-02-11T15:15:42.098645Z"}}},{"cell_type":"code","source":"!pip install pyunpack\n!pip install patool\n!pip install py7zr\n!pip install sounddevice\n!pip install noisereduce\n!pip install librosa\n! pip install python_speech_features\n! pip install tensorflow==2.4\n! pip install malaya_speech\n! pip install webrtcvad","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:25:27.353518Z","iopub.execute_input":"2022-02-13T20:25:27.353910Z","iopub.status.idle":"2022-02-13T20:27:18.441710Z","shell.execute_reply.started":"2022-02-13T20:25:27.353827Z","shell.execute_reply":"2022-02-13T20:27:18.440227Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\nfrom py7zr import unpack_7zarchive\nimport shutil\nimport warnings\nwarnings.filterwarnings(\"ignore\")\n\nimport matplotlib.pyplot as plt\nimport numpy as np\n\nimport librosa\nimport IPython.display as ipd\nfrom scipy.io import wavfile\n\nimport noisereduce as nr\nimport tensorflow \nfrom malaya_speech import Pipeline\n\nimport malaya_speech\nimport os\n\nfrom python_speech_features import mfcc\n\nfrom sklearn.preprocessing import LabelEncoder\nimport seaborn as sn","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:27:18.448783Z","iopub.execute_input":"2022-02-13T20:27:18.449256Z","iopub.status.idle":"2022-02-13T20:27:23.227613Z","shell.execute_reply.started":"2022-02-13T20:27:18.449212Z","shell.execute_reply":"2022-02-13T20:27:23.226444Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"shutil.register_unpack_format('7zip', ['.7z'], unpack_7zarchive)\nshutil.unpack_archive('/kaggle/input/tensorflow-speech-recognition-challenge/train.7z', '/kaggle/working/tensorflow-speech-recognition-challenge/train/')","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:27:23.229601Z","iopub.execute_input":"2022-02-13T20:27:23.229955Z","iopub.status.idle":"2022-02-13T20:29:56.175038Z","shell.execute_reply.started":"2022-02-13T20:27:23.229919Z","shell.execute_reply":"2022-02-13T20:29:56.174145Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# from pyunpack import Archive\n# import shutil\n# if not os.path.exists('/kaggle/working/tensorflow-speech-recognition-challenge/train/'):\n#     os.makedirs('/kaggle/working/tensorflow-speech-recognition-challenge/train/')\n# Archive('/kaggle/input/tensorflow-speech-recognition-challenge/train.7z').extractall('/kaggle/working/tensorflow-speech-recognition-challenge/train/')\n\n#for dirname, _, filenames in os.walk('/kaggle/working/tensorflow-speech-recognition-challenge/train/train/audio'):\n #   for filename in filename[:5]:\n  #      print(os.path.join(dirname, filename))","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:29:56.176730Z","iopub.execute_input":"2022-02-13T20:29:56.177056Z","iopub.status.idle":"2022-02-13T20:29:56.181791Z","shell.execute_reply.started":"2022-02-13T20:29:56.177020Z","shell.execute_reply":"2022-02-13T20:29:56.181008Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# <center> Implementing the Speech Recognition Model in Python\n**Dataset used for our Speech Recognition Project**\n\n__You can download the dataset from__ [here](https://www.kaggle.com/c/tensorflow-speech-recognition-challenge).\n    \nTensorFlow recently released the Speech Commands Datasets. It includes 65,000 one-second long utterances of 30 short words, by thousands of different people. We’ll build a speech recognition system that understands simple spoken commands. <br>    ","metadata":{}},{"cell_type":"markdown","source":"**Data Exploration and Visualization**\n\nData Exploration and Visualization helps us to understand the data as well as pre-processing steps in a better way. \n\n**Visualization of Audio signal in time series domain**\n\nNow, we’ll visualize the audio signal in the time series domain:","metadata":{}},{"cell_type":"code","source":"train_audio_path = '/kaggle/working/tensorflow-speech-recognition-challenge/train/train/audio/'","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:29:56.183240Z","iopub.execute_input":"2022-02-13T20:29:56.183875Z","iopub.status.idle":"2022-02-13T20:29:56.191083Z","shell.execute_reply.started":"2022-02-13T20:29:56.183831Z","shell.execute_reply":"2022-02-13T20:29:56.190250Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"**Accessing each file in data**","metadata":{}},{"cell_type":"code","source":"#!apt-get install -y p7zip-full\n#!7z x ../input/tensorflow-speech-recognition-challenge/train.7z","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:29:56.192871Z","iopub.execute_input":"2022-02-13T20:29:56.193568Z","iopub.status.idle":"2022-02-13T20:29:56.199690Z","shell.execute_reply.started":"2022-02-13T20:29:56.193531Z","shell.execute_reply":"2022-02-13T20:29:56.198845Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"samples, sample_rate = librosa.load(train_audio_path+'on/5a3712c9_nohash_1.wav', sr = 16000)\nfig = plt.figure(figsize=(14, 8))\nax1 = fig.add_subplot(211)\nax1.set_title('Raw wave of ' + '../input/train/audio/on/0a7c2a8d_nohash_0.wav')\nax1.set_xlabel('time')\nax1.set_ylabel('Amplitude')\nax1.plot(np.linspace(0, sample_rate/len(samples), sample_rate), samples)","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:29:56.200811Z","iopub.execute_input":"2022-02-13T20:29:56.201173Z","iopub.status.idle":"2022-02-13T20:29:56.383638Z","shell.execute_reply.started":"2022-02-13T20:29:56.201138Z","shell.execute_reply":"2022-02-13T20:29:56.382790Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"**Sampling rate **\n\nLet us now look at the sampling rate of the audio signals","metadata":{}},{"cell_type":"code","source":"ipd.Audio(samples, rate=sample_rate)","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:29:56.386217Z","iopub.execute_input":"2022-02-13T20:29:56.386717Z","iopub.status.idle":"2022-02-13T20:29:56.395715Z","shell.execute_reply.started":"2022-02-13T20:29:56.386679Z","shell.execute_reply":"2022-02-13T20:29:56.393656Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(sample_rate)\nsig1=samples\nfs=sample_rate\nsr=fs","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:29:56.397700Z","iopub.execute_input":"2022-02-13T20:29:56.398261Z","iopub.status.idle":"2022-02-13T20:29:56.403672Z","shell.execute_reply.started":"2022-02-13T20:29:56.398224Z","shell.execute_reply":"2022-02-13T20:29:56.402690Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"time = np.linspace(0, len(sig1 - 1) / fs, len(sig1 - 1))\nreduced_noise1 = nr.reduce_noise(y=sig1, sr=fs,stationary=True)\nplt.plot(time, reduced_noise1)  # plot in seconds\n#reduced_noise2 = nr.reduce_noise(y=sig2, sr=fs,stationary=True)\n#plt.plot(time, reduced_noise2)  # plot in seconds\n#plt.title(\"Voice Signal\")\nplt.xlabel(\"Time [seconds]\")\nplt.ylabel(\"Voice amplitude\")\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:29:56.404869Z","iopub.execute_input":"2022-02-13T20:29:56.405503Z","iopub.status.idle":"2022-02-13T20:29:57.221171Z","shell.execute_reply.started":"2022-02-13T20:29:56.405466Z","shell.execute_reply":"2022-02-13T20:29:57.220289Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"ipd.Audio(reduced_noise1, rate=sample_rate)","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:29:57.222654Z","iopub.execute_input":"2022-02-13T20:29:57.223261Z","iopub.status.idle":"2022-02-13T20:29:57.235331Z","shell.execute_reply.started":"2022-02-13T20:29:57.223223Z","shell.execute_reply":"2022-02-13T20:29:57.234585Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#Silence Removal\nvad = malaya_speech.vad.webrtc()\ny=reduced_noise1\ny_= malaya_speech.resample(y, sr, 16000)\ny_ = malaya_speech.astype.float_to_int(y_)\nframes = malaya_speech.generator.frames(y, 30, sr)\nframes_ = list(malaya_speech.generator.frames(y_, 30, 16000, append_ending_trail = False))\nframes_webrtc = [(frames[no], vad(frame)) for no, frame in enumerate(frames_)]\ny_ = malaya_speech.combine.without_silent(frames_webrtc)\ny_","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:29:57.237061Z","iopub.execute_input":"2022-02-13T20:29:57.240714Z","iopub.status.idle":"2022-02-13T20:29:57.505189Z","shell.execute_reply.started":"2022-02-13T20:29:57.240676Z","shell.execute_reply":"2022-02-13T20:29:57.504398Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"ipd.Audio(y_, rate = sr )","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:29:57.509066Z","iopub.execute_input":"2022-02-13T20:29:57.511228Z","iopub.status.idle":"2022-02-13T20:29:57.521441Z","shell.execute_reply.started":"2022-02-13T20:29:57.511169Z","shell.execute_reply":"2022-02-13T20:29:57.520594Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"zero = np.zeros((1*sr-y_.shape[0]))\nsignal = np.concatenate((y_,zero))\nsignal.shape\ntime = np.linspace(0, len(signal - 1) / fs, len(signal - 1))","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:29:57.525406Z","iopub.execute_input":"2022-02-13T20:29:57.527331Z","iopub.status.idle":"2022-02-13T20:29:57.533877Z","shell.execute_reply.started":"2022-02-13T20:29:57.527294Z","shell.execute_reply":"2022-02-13T20:29:57.533213Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"plt.plot(time,signal)","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:29:57.537310Z","iopub.execute_input":"2022-02-13T20:29:57.539276Z","iopub.status.idle":"2022-02-13T20:29:57.697938Z","shell.execute_reply.started":"2022-02-13T20:29:57.539240Z","shell.execute_reply":"2022-02-13T20:29:57.697037Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"labels=os.listdir(train_audio_path)","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:29:57.699323Z","iopub.execute_input":"2022-02-13T20:29:57.699893Z","iopub.status.idle":"2022-02-13T20:29:57.704437Z","shell.execute_reply.started":"2022-02-13T20:29:57.699841Z","shell.execute_reply":"2022-02-13T20:29:57.703341Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#find count of each label and plot bar graph\nno_of_recordings=[]\nfor label in labels:\n    waves = [f for f in os.listdir(train_audio_path + '/'+ label) if f.endswith('.wav')]\n    no_of_recordings.append(len(waves))\n    \n#plot\nplt.figure(figsize=(30,5))\nindex = np.arange(len(labels))\nplt.bar(index, no_of_recordings)\nplt.xlabel('Commands', fontsize=12)\nplt.ylabel('No of recordings', fontsize=12)\nplt.xticks(index, labels, fontsize=15, rotation=60)\nplt.title('No. of recordings for each command')\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:29:57.706061Z","iopub.execute_input":"2022-02-13T20:29:57.706694Z","iopub.status.idle":"2022-02-13T20:29:58.076335Z","shell.execute_reply.started":"2022-02-13T20:29:57.706654Z","shell.execute_reply":"2022-02-13T20:29:58.075524Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Words used","metadata":{}},{"cell_type":"code","source":"labels=[\"zero\",\"one\",\"two\",\"three\",\"four\",\"five\",\"six\",\"seven\",\"eight\",\"nine\"]","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:29:58.077755Z","iopub.execute_input":"2022-02-13T20:29:58.078140Z","iopub.status.idle":"2022-02-13T20:29:58.083910Z","shell.execute_reply.started":"2022-02-13T20:29:58.078102Z","shell.execute_reply":"2022-02-13T20:29:58.082899Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## **Preprocessing the audio waves**\n\nlet us read the audio waves and use the below-preprocessing steps :\n\n* Noise Reduction\n* Silence Removal\n* Extracting MFCCs\n\nLet us define these preprocessing steps in the below code snippet:","metadata":{}},{"cell_type":"code","source":"sr=16000\nvad = malaya_speech.vad.webrtc()\nall_wave = []\nall_label = []\nfor label in labels:\n    print(label)\n    waves = [f for f in os.listdir(train_audio_path + '/'+ label) if f.endswith('.wav')]\n    for wav in waves:\n        samples, sample_rate = librosa.load(train_audio_path + '/' + label + '/' + wav, sr = 16000)\n        samples = nr.reduce_noise(y=samples, sr=sr,stationary=True)\n        y_= malaya_speech.resample(samples, sr, 16000)\n        y_ = malaya_speech.astype.float_to_int(y_)\n        frames = malaya_speech.generator.frames(samples, 30, sr)\n        frames_ = list(malaya_speech.generator.frames(y_, 30, 16000, append_ending_trail = False))\n        frames_webrtc = [(frames[no], vad(frame)) for no, frame in enumerate(frames_)]\n        y_ = malaya_speech.combine.without_silent(frames_webrtc)\n        zero = np.zeros(((1*sr+4000)-y_.shape[0]))\n        signal = np.concatenate((y_,zero))\n        all_wave.append(signal)\n        all_label.append(label)","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:29:58.085316Z","iopub.execute_input":"2022-02-13T20:29:58.085896Z","iopub.status.idle":"2022-02-13T20:41:25.257890Z","shell.execute_reply.started":"2022-02-13T20:29:58.085838Z","shell.execute_reply":"2022-02-13T20:41:25.257034Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(np.array(all_wave).shape)\nprint(np.array(all_label).shape)\ntime = np.linspace(0, len(signal - 1) / fs, len(signal - 1))\nplt.plot(time,np.array(all_wave)[2000,:])\nprint(np.array(all_label)[2000])\nipd.Audio(np.array(all_wave)[2000,:], rate = sr )","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:41:25.259275Z","iopub.execute_input":"2022-02-13T20:41:25.259606Z","iopub.status.idle":"2022-02-13T20:41:36.022996Z","shell.execute_reply.started":"2022-02-13T20:41:25.259571Z","shell.execute_reply":"2022-02-13T20:41:36.022072Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"all_mfcc=[]\nfor wave in all_wave:\n    i=0\n    mfcc_feat = mfcc(wave , fs, winlen=256/fs, winstep=256/(2*fs), numcep=13, nfilt=26, nfft=256,\n                 lowfreq=0, highfreq=fs/2, preemph=0.97, ceplifter=22, appendEnergy=True, winfunc=np.hamming)\n    mfcc_feat= np.transpose(mfcc_feat)\n    all_mfcc.append(mfcc_feat)\n    ","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:41:36.024301Z","iopub.execute_input":"2022-02-13T20:41:36.024647Z","iopub.status.idle":"2022-02-13T20:42:47.330550Z","shell.execute_reply.started":"2022-02-13T20:41:36.024608Z","shell.execute_reply":"2022-02-13T20:42:47.329322Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(np.array(all_mfcc).shape)\nprint(np.array(all_label).shape)\nd1=np.array(all_mfcc).shape[1]\nd2=np.array(all_mfcc).shape[2]\nd=d1*d2\nprint(d)","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:42:47.331948Z","iopub.execute_input":"2022-02-13T20:42:47.332287Z","iopub.status.idle":"2022-02-13T20:42:47.953773Z","shell.execute_reply.started":"2022-02-13T20:42:47.332253Z","shell.execute_reply":"2022-02-13T20:42:47.952837Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"op_mfcc=np.array(all_mfcc)\nop_mfcc=op_mfcc.reshape(np.array(all_mfcc).shape[0],-1)\nop_mfcc.shape","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:42:47.957764Z","iopub.execute_input":"2022-02-13T20:42:47.958018Z","iopub.status.idle":"2022-02-13T20:42:48.270823Z","shell.execute_reply.started":"2022-02-13T20:42:47.957992Z","shell.execute_reply":"2022-02-13T20:42:48.269837Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#all_label = all_label.tolist()\n\nle = LabelEncoder()\ny=le.fit_transform(all_label)\nclasses= list(le.classes_)","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:42:48.272945Z","iopub.execute_input":"2022-02-13T20:42:48.273331Z","iopub.status.idle":"2022-02-13T20:42:48.285217Z","shell.execute_reply.started":"2022-02-13T20:42:48.273293Z","shell.execute_reply":"2022-02-13T20:42:48.284457Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## **Model based on ANN** ","metadata":{}},{"cell_type":"code","source":"! pip install --upgrade tensorflow\n! pip install --upgrade tensorflow-gpu\n! pip install keras==2.3.1","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:42:48.286531Z","iopub.execute_input":"2022-02-13T20:42:48.287136Z","iopub.status.idle":"2022-02-13T20:44:51.048398Z","shell.execute_reply.started":"2022-02-13T20:42:48.287094Z","shell.execute_reply":"2022-02-13T20:44:51.047424Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"from keras.optimizers import SGD\nfrom keras.constraints import maxnorm\nfrom tensorflow.keras import Sequential\nfrom tensorflow.keras.layers import Conv2D, Flatten, Dense,Dropout\nfrom keras.callbacks import EarlyStopping, ModelCheckpoint","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:44:51.051026Z","iopub.execute_input":"2022-02-13T20:44:51.051405Z","iopub.status.idle":"2022-02-13T20:44:51.105688Z","shell.execute_reply.started":"2022-02-13T20:44:51.051366Z","shell.execute_reply":"2022-02-13T20:44:51.104814Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"y=tensorflow.keras.utils.to_categorical(y, num_classes=len(labels), dtype='float32')\ny.shape","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:44:51.107075Z","iopub.execute_input":"2022-02-13T20:44:51.107680Z","iopub.status.idle":"2022-02-13T20:44:51.116586Z","shell.execute_reply.started":"2022-02-13T20:44:51.107639Z","shell.execute_reply":"2022-02-13T20:44:51.115726Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"from sklearn.model_selection import train_test_split\nx_tr, x_val, y_tr, y_val= train_test_split(op_mfcc,np.array(y),stratify=y,test_size = 0.2,random_state=777,shuffle=True)","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:44:51.118345Z","iopub.execute_input":"2022-02-13T20:44:51.118599Z","iopub.status.idle":"2022-02-13T20:44:51.839948Z","shell.execute_reply.started":"2022-02-13T20:44:51.118575Z","shell.execute_reply":"2022-02-13T20:44:51.839086Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(x_tr.shape)\nprint(y_tr.shape)\nprint(x_val.shape)\nprint(y_val.shape)","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:44:51.841174Z","iopub.execute_input":"2022-02-13T20:44:51.841534Z","iopub.status.idle":"2022-02-13T20:44:51.846962Z","shell.execute_reply.started":"2022-02-13T20:44:51.841500Z","shell.execute_reply":"2022-02-13T20:44:51.846137Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### **Model Architecture**","metadata":{}},{"cell_type":"code","source":"#Model Architecture\nmodel = Sequential()\nmodel.add(Dense(100, activation='relu', input_shape=(d,), kernel_constraint=maxnorm(3)))\nmodel.add(Dropout(0.5))\nmodel.add(Dense(80, activation='relu', kernel_constraint=maxnorm(3)))\nmodel.add(Dropout(0.5))\nmodel.add(Dense(len(classes), activation='softmax' , kernel_constraint=maxnorm(3)))","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:44:51.848172Z","iopub.execute_input":"2022-02-13T20:44:51.848667Z","iopub.status.idle":"2022-02-13T20:44:54.233611Z","shell.execute_reply.started":"2022-02-13T20:44:51.848629Z","shell.execute_reply":"2022-02-13T20:44:54.232675Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"tensorflow.keras.utils.plot_model(model, 'model.png',show_shapes=True)","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:44:54.234804Z","iopub.execute_input":"2022-02-13T20:44:54.235139Z","iopub.status.idle":"2022-02-13T20:44:54.836019Z","shell.execute_reply.started":"2022-02-13T20:44:54.235095Z","shell.execute_reply":"2022-02-13T20:44:54.835096Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model.compile(loss='categorical_crossentropy',optimizer='adamax',metrics=['accuracy'])\n","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:44:54.837755Z","iopub.execute_input":"2022-02-13T20:44:54.838341Z","iopub.status.idle":"2022-02-13T20:44:54.857340Z","shell.execute_reply.started":"2022-02-13T20:44:54.838298Z","shell.execute_reply":"2022-02-13T20:44:54.856546Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"es = EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=10, min_delta=0.0001) \nmc = ModelCheckpoint('best_model.hdf5', monitor='val_acc', verbose=1, save_best_only=True, mode='max')","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:44:54.859288Z","iopub.execute_input":"2022-02-13T20:44:54.860128Z","iopub.status.idle":"2022-02-13T20:44:54.865565Z","shell.execute_reply.started":"2022-02-13T20:44:54.860087Z","shell.execute_reply":"2022-02-13T20:44:54.864765Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#18932\n300*65","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:44:54.866913Z","iopub.execute_input":"2022-02-13T20:44:54.867322Z","iopub.status.idle":"2022-02-13T20:44:54.879164Z","shell.execute_reply.started":"2022-02-13T20:44:54.867288Z","shell.execute_reply":"2022-02-13T20:44:54.878210Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"history=model.fit(x_tr, y_tr,validation_data=(x_val,y_val), epochs=300, batch_size=65)","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:44:54.880556Z","iopub.execute_input":"2022-02-13T20:44:54.881061Z","iopub.status.idle":"2022-02-13T20:49:53.016435Z","shell.execute_reply.started":"2022-02-13T20:44:54.881025Z","shell.execute_reply":"2022-02-13T20:49:53.015526Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_score = model.evaluate(x_tr, y_tr, batch_size=12)\nprint(train_score)\n\nprint('----------------Training Complete-----------------')\n\ntest_score = model.evaluate(x_val, y_val, batch_size = 12)\nprint(test_score)","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:49:53.017927Z","iopub.execute_input":"2022-02-13T20:49:53.018329Z","iopub.status.idle":"2022-02-13T20:49:57.104560Z","shell.execute_reply.started":"2022-02-13T20:49:53.018287Z","shell.execute_reply":"2022-02-13T20:49:57.103751Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"history.history.keys()","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:49:57.107258Z","iopub.execute_input":"2022-02-13T20:49:57.107520Z","iopub.status.idle":"2022-02-13T20:49:57.116159Z","shell.execute_reply.started":"2022-02-13T20:49:57.107493Z","shell.execute_reply":"2022-02-13T20:49:57.115292Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"from matplotlib import pyplot\npyplot.plot(history.history['loss'], label='train')\npyplot.plot(history.history['val_loss'], label='test')\npyplot.legend()\npyplot.show()","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:49:57.117474Z","iopub.execute_input":"2022-02-13T20:49:57.117827Z","iopub.status.idle":"2022-02-13T20:49:57.271059Z","shell.execute_reply.started":"2022-02-13T20:49:57.117791Z","shell.execute_reply":"2022-02-13T20:49:57.270179Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"plt.plot(history.history['accuracy'])\nplt.plot(history.history['val_accuracy'])\nplt.title('model accuracy')\nplt.ylabel('accuracy')\nplt.xlabel('epoch')\nplt.legend(['train', 'val'], loc='upper left')\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:49:57.272387Z","iopub.execute_input":"2022-02-13T20:49:57.272886Z","iopub.status.idle":"2022-02-13T20:49:57.410906Z","shell.execute_reply.started":"2022-02-13T20:49:57.272846Z","shell.execute_reply":"2022-02-13T20:49:57.410069Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"y_predict=model.predict(x_val)\nconf_mat=tensorflow.math.confusion_matrix(np.argmax(y_val,axis=1) , np.argmax(y_predict,axis=1))","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:49:57.412143Z","iopub.execute_input":"2022-02-13T20:49:57.412670Z","iopub.status.idle":"2022-02-13T20:49:58.086537Z","shell.execute_reply.started":"2022-02-13T20:49:57.412629Z","shell.execute_reply":"2022-02-13T20:49:58.085613Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"df_cm = pd.DataFrame(np.array(conf_mat), index = [i for i in classes],\n                  columns = [i for i in classes])\nplt.figure(figsize = (13,7))\nax = sn.heatmap(df_cm, annot=True)\nplt.title(\"Confusion Matrix\", fontsize=20)\nplt.ylabel(\"True Class\"     , fontsize=20)\nplt.xlabel(\"Predicted Class\", fontsize=20)\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:49:58.087916Z","iopub.execute_input":"2022-02-13T20:49:58.088285Z","iopub.status.idle":"2022-02-13T20:49:58.697889Z","shell.execute_reply.started":"2022-02-13T20:49:58.088248Z","shell.execute_reply":"2022-02-13T20:49:58.697040Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"x_val[1].shape\nmodel.predict(x_val[1].reshape((1,d)))","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:49:58.699107Z","iopub.execute_input":"2022-02-13T20:49:58.699619Z","iopub.status.idle":"2022-02-13T20:49:58.981722Z","shell.execute_reply.started":"2022-02-13T20:49:58.699579Z","shell.execute_reply":"2022-02-13T20:49:58.980893Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Define the function that predicts text for the given audio:","metadata":{}},{"cell_type":"code","source":"def predict(audio):\n    print(samples.shape)\n    prob=model.predict(audio)\n    index=np.argmax(prob[0])\n    return classes[index]","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:49:58.983016Z","iopub.execute_input":"2022-02-13T20:49:58.983360Z","iopub.status.idle":"2022-02-13T20:49:58.990206Z","shell.execute_reply.started":"2022-02-13T20:49:58.983323Z","shell.execute_reply":"2022-02-13T20:49:58.989443Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Prediction time! Make predictions on the validation data:","metadata":{}},{"cell_type":"code","source":"import random\nindex=random.randint(0,len(x_val)-1)\nprint(index)\nsamples=x_val[index]\nprint(\"Audio:\",classes[np.argmax(y_val[index])])\n#ipd.Audio(np.array(all_wave)[index,:], rate=16000)","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:49:58.991566Z","iopub.execute_input":"2022-02-13T20:49:58.991928Z","iopub.status.idle":"2022-02-13T20:49:59.000639Z","shell.execute_reply.started":"2022-02-13T20:49:58.991892Z","shell.execute_reply":"2022-02-13T20:49:58.999571Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(\"Text:\",predict(samples.reshape(1,d)))","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:49:59.002104Z","iopub.execute_input":"2022-02-13T20:49:59.002644Z","iopub.status.idle":"2022-02-13T20:49:59.043709Z","shell.execute_reply.started":"2022-02-13T20:49:59.002608Z","shell.execute_reply":"2022-02-13T20:49:59.042890Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"from keras.models import load_model\nmodel.save(\"MME.h5\")","metadata":{"execution":{"iopub.status.busy":"2022-02-13T20:49:59.046392Z","iopub.execute_input":"2022-02-13T20:49:59.046645Z","iopub.status.idle":"2022-02-13T20:49:59.090584Z","shell.execute_reply.started":"2022-02-13T20:49:59.046620Z","shell.execute_reply":"2022-02-13T20:49:59.089715Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]}]}