File size: 4,559 Bytes
e00e2b3
 
 
 
 
 
 
 
f490873
e00e2b3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
13f320b
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
137
138
139
140
141
import torch
import torch.nn as nn
import torchaudio
import sounddevice as sd
import scipy.io.wavfile as wav
import numpy as np
import os

MODEL_PATH = "model_best.pth" 
DURATION = 5  
VOCAB_STR = "_abcçdefgğhıijklmnoöprsştuüvyzqwx "

class ResCNNBlock(nn.Module):
    def __init__(self, in_channels, out_channels):
        super(ResCNNBlock, self).__init__()
        self.conv1 = nn.Conv2d(in_channels, out_channels, kernel_size=3, stride=1, padding=1)
        self.bn1 = nn.BatchNorm2d(out_channels)
        self.act1 = nn.GELU()
        self.conv2 = nn.Conv2d(out_channels, out_channels, kernel_size=3, stride=1, padding=1)
        self.bn2 = nn.BatchNorm2d(out_channels)
        self.act2 = nn.GELU()

    def forward(self, x):
        residual = x
        out = self.conv1(x)
        out = self.bn1(out)
        out = self.act1(out)
        out = self.conv2(out)
        out = self.bn2(out)
        out = self.act2(out)
        if residual.shape[1] == out.shape[1]:
            out += residual
        return out

class DeepSpeechModel(nn.Module):
    def __init__(self, num_classes):
        super(DeepSpeechModel, self).__init__()
        self.cnn = nn.Sequential(
            nn.Conv2d(1, 32, 3, stride=2, padding=1), nn.GELU(),
            ResCNNBlock(32, 32),
            ResCNNBlock(32, 32),
            ResCNNBlock(32, 64),
            ResCNNBlock(64, 64),
            nn.Dropout(0.1)
        )
        rnn_input_size = 64 * 64 
        self.dense = nn.Linear(rnn_input_size, 1024)
        self.layer_norm = nn.LayerNorm(1024)
        self.rnn = nn.LSTM(input_size=1024, hidden_size=512, num_layers=4, 
                           batch_first=True, bidirectional=True, dropout=0.3)
        self.classifier = nn.Linear(512*2, num_classes)

    def forward(self, x):
        x = self.cnn(x) 
        b, c, t, f = x.shape
        x = x.permute(0, 2, 1, 3).contiguous().view(b, t, c*f)
        x = self.dense(x)
        x = self.layer_norm(x)
        x, _ = self.rnn(x)
        x = self.classifier(x)
        return x

def greedy_decoder(output, vocab):
    arg_maxes = torch.argmax(output, dim=2).squeeze().tolist()
    decoded_chars = []
    prev_index = -1
    id_to_char = {i: char for i, char in enumerate(vocab)}
    
    for index in arg_maxes:
        if index != prev_index:
            if index != 0: 
                char = id_to_char.get(index, "")
                decoded_chars.append(char)
        prev_index = index
        
    return "".join(decoded_chars)

def record_audio(duration, fs, filename):
    print(f"\nRECORDING... ({duration} s)")
    try:
        recording = sd.rec(int(duration * fs), samplerate=fs, channels=1, dtype='float32')
        sd.wait()
        print("Recording finished.")
        wav.write(filename, fs, (recording * 32767).astype(np.int16))
    except Exception as e:
        print(f"Recording Error: {e}")

def predict(audio_path):
    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    
    vocab = list(VOCAB_STR)
    
    if not os.path.exists(MODEL_PATH):
        print(f"ERROR: {MODEL_PATH} not found!")
        return

    checkpoint = torch.load(MODEL_PATH, map_location=device,weights_only=True)
    saved_vocab_size = checkpoint['classifier.bias'].shape[0]
    
    if len(vocab) != saved_vocab_size:
        while len(vocab) < saved_vocab_size:
            vocab.append("?") 

    model = DeepSpeechModel(num_classes=saved_vocab_size).to(device)
    model.load_state_dict(checkpoint)
    model.eval()

    waveform, sr = torchaudio.load(audio_path)
    if sr != 16000: 
        waveform = torchaudio.transforms.Resample(sr, 16000)(waveform)
    if waveform.shape[0] > 1: 
        waveform = torch.mean(waveform, dim=0, keepdim=True)

    mel_transform = torchaudio.transforms.MelSpectrogram(
        sample_rate=16000, n_mels=128, n_fft=1024, hop_length=256
    ).to(device)
    
    spec = mel_transform(waveform.to(device))
    spec = torch.log(spec + 1e-9)
    if spec.dim() == 2: spec = spec.unsqueeze(0)
    spec = spec.unsqueeze(1)
    spec = spec.permute(0, 1, 3, 2) 

    with torch.no_grad():
        output = model(spec)
        
    text = greedy_decoder(output, vocab)
    text = text.replace("_", " ") 
    
    print("-" * 40)
    print(f"RECOGNIZED: {text}")
    print("-" * 40)

if __name__ == "__main__":
    temp_file = "live_final.wav"
    while True:
        user_input = input("Press Enter to record, q to exit: ")
        if user_input.lower() == "q":
            break
        record_audio(DURATION, 16000, temp_file)
        predict(temp_file)