Skip to content

delta_feat

Author: Heli Qi Affiliation: NAIST Date: 2022.07

DeltaFeature

Bases: Module

Source code in speechain/module/frontend/delta_feat.py 
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
class DeltaFeature(Module):
    """"""

    def module_init(self, delta_order: int = 1, delta_N: int = 2):
        """
        modified from https://github.com/jameslyons/python_speech_features/blob/master/python_speech_features/base.py#L195

        Args:
            delta_order:
            delta_N:

        Returns:

        """
        assert isinstance(delta_order, int) and delta_order in [
            1,
            2,
        ], f"delta_order must be an integer of 1 or 2, but got delta_order={delta_order}"
        self.delta_order = delta_order

        # initialze the filters for extracting delta features
        assert (
            isinstance(delta_N, int) and delta_N >= 1
        ), "delta_N must be a positive integer equal to or larger than 1"
        self.delta_N = delta_N
        _filter_weights = torch.arange(-delta_N, delta_N + 1) / (
            2 * sum([i**2 for i in range(1, delta_N + 1)])
        )
        _delta_filters = torch.nn.Conv2d(
            in_channels=1,
            out_channels=1,
            kernel_size=(2 * delta_N + 1, 1),
            padding=(delta_N, 0),
            bias=False,
        )
        _delta_filters.weight = torch.nn.Parameter(
            _filter_weights.reshape(1, 1, -1, 1), requires_grad=False
        )

        # move the weight from _parameters to _buffers
        _para_keys = list(_delta_filters._parameters.keys())
        for name in _para_keys:
            _delta_filters._buffers[name] = _delta_filters._parameters.pop(name)
        self.delta_filters = _delta_filters

    def forward(self, feat: torch.Tensor, feat_len: torch.Tensor):
        """

        Args:
            feat:
            feat_len:

        Returns:

        """

        # first-order derivative
        feat_stack = [feat]
        delta_feat = self.delta_filters(feat.unsqueeze(1)).squeeze(1)
        feat_stack.append(delta_feat)

        # (Optional) second-order derivative
        if self.delta_order == 2:
            delta2_feat = self.delta_filters(delta_feat.unsqueeze(1)).squeeze(1)
            feat_stack.append(delta2_feat)

        # combine the original features with all delta features
        feat = torch.cat(feat_stack, dim=-1)

        return feat, feat_len

    def __repr__(self) -> str:
        return (
            f"{self.__class__.__name__}(\n"
            f"delta_order={self.delta_order}, "
            f"delta_N={self.delta_N}"
            f"\n)"
        )

forward(feat, feat_len)

Parameters:

Name Type Description Default
feat Tensor
required
feat_len Tensor
required

Returns:

Source code in speechain/module/frontend/delta_feat.py 
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
def forward(self, feat: torch.Tensor, feat_len: torch.Tensor):
    """

    Args:
        feat:
        feat_len:

    Returns:

    """

    # first-order derivative
    feat_stack = [feat]
    delta_feat = self.delta_filters(feat.unsqueeze(1)).squeeze(1)
    feat_stack.append(delta_feat)

    # (Optional) second-order derivative
    if self.delta_order == 2:
        delta2_feat = self.delta_filters(delta_feat.unsqueeze(1)).squeeze(1)
        feat_stack.append(delta2_feat)

    # combine the original features with all delta features
    feat = torch.cat(feat_stack, dim=-1)

    return feat, feat_len

module_init(delta_order=1, delta_N=2)

modified from https://github.com/jameslyons/python_speech_features/blob/master/python_speech_features/base.py#L195

Parameters:

Name Type Description Default
delta_order int
1
delta_N int
2

Returns:

Source code in speechain/module/frontend/delta_feat.py 
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
def module_init(self, delta_order: int = 1, delta_N: int = 2):
    """
    modified from https://github.com/jameslyons/python_speech_features/blob/master/python_speech_features/base.py#L195

    Args:
        delta_order:
        delta_N:

    Returns:

    """
    assert isinstance(delta_order, int) and delta_order in [
        1,
        2,
    ], f"delta_order must be an integer of 1 or 2, but got delta_order={delta_order}"
    self.delta_order = delta_order

    # initialze the filters for extracting delta features
    assert (
        isinstance(delta_N, int) and delta_N >= 1
    ), "delta_N must be a positive integer equal to or larger than 1"
    self.delta_N = delta_N
    _filter_weights = torch.arange(-delta_N, delta_N + 1) / (
        2 * sum([i**2 for i in range(1, delta_N + 1)])
    )
    _delta_filters = torch.nn.Conv2d(
        in_channels=1,
        out_channels=1,
        kernel_size=(2 * delta_N + 1, 1),
        padding=(delta_N, 0),
        bias=False,
    )
    _delta_filters.weight = torch.nn.Parameter(
        _filter_weights.reshape(1, 1, -1, 1), requires_grad=False
    )

    # move the weight from _parameters to _buffers
    _para_keys = list(_delta_filters._parameters.keys())
    for name in _para_keys:
        _delta_filters._buffers[name] = _delta_filters._parameters.pop(name)
    self.delta_filters = _delta_filters