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
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
/*!
example of using the mel create to build up a filter matrix that can then
be used to
repeatedly and efficiently (in-place)
transform a huge (2048) power spectrum vector
to a mel filtered vector of lower dimension (100).

```
extern crate mel;

extern crate nalgebra;
use nalgebra::{DMatrix, DVector};

fn main() {
    let sample_rate = 44100;
    let window_size = 4096;
    let power_spectrum_size = window_size / 2;
    let filter_count = 100;

    // build up the mel filter matrix
    let mut mel_filter_matrix =
        DMatrix::<f64>::zeros(filter_count, power_spectrum_size);
    for (row, col, coefficient) in mel::enumerate_mel_scaling_matrix(
        sample_rate,
        window_size,
        power_spectrum_size,
        filter_count,
    ) {
        mel_filter_matrix[(row, col)] = coefficient;
    }

    let power_spectrum_vector =
        DVector::<f64>::from_element(power_spectrum_size, 1.);

    let mut mel_vector =
        DVector::<f64>::zeros(filter_count);

    // mel_vector <- mel_filter_matrix * power_spectrum_vector
    mel_filter_matrix.mul_to(&power_spectrum_vector, &mut mel_vector);

    // mel_vector now contains a mel filtered version
    // of the power_spectrum_vector!
}
```
*/

extern crate num;
use num::{Float, FromPrimitive};

extern crate hertz;
extern crate apodize;

extern crate itertools_num;
use itertools_num::linspace;

macro_rules! f64_from_usize {
    ($val:expr) => {
        <f64>::from_usize($val)
            .expect("type `f64` can't represent a specific value of type `usize` on this architecture.");
    }
}

macro_rules! usize_from_f64 {
    ($val:expr) => {
        <usize>::from_f64($val)
            .expect("type `usize` can't represent a specific value of type `f64` on this architecture.");
    }
}

#[inline]
pub fn hertz_from_mel(mel: f64) -> f64 {
    700. * ((10.).powf(mel / 2595.) - 1.)
}

#[inline]
pub fn mel_from_hertz(hertz: f64) -> f64 {
    2595. * (1. + hertz / 700.).log10()
}

/// col index changes fastest
pub struct MelScalingMatrixEnumerator<WindowIter> {
    // parameters

    /// equivalent to col count
    input_size: usize,
    /// equivalent to row count
    output_size: usize,

    mel_from_hertz: fn(f64) -> f64,
    hertz_from_mel: fn(f64) -> f64,

    max_hertz: f64,

    window_function: fn(usize) -> WindowIter,

    // state

    row_index: usize,
    col_index: usize,

    start_mels_iter: itertools_num::Linspace<f64>,
    end_mels_iter: itertools_num::Linspace<f64>,

    // this gets set anew for every row
    window_start: usize,
    window_size: usize,
    window_iter: WindowIter
}

impl<WindowIter> MelScalingMatrixEnumerator<WindowIter>
    where WindowIter: Iterator<Item=f64>
{
    #[inline]
    pub fn is_done(&self) -> bool {
        self.is_after_last_row() && self.is_after_last_col()
    }

    #[inline]
    pub fn is_after_last_row(&self) -> bool {
        self.output_size <= self.row_index
    }

    #[inline]
    pub fn is_after_last_col(&self) -> bool {
        self.input_size <= self.col_index
    }

    #[inline]
    pub fn index(&self) -> usize {
        self.row_index * self.input_size + self.col_index
    }

    #[inline]
    pub fn remaining(&self) -> usize {
        // TODO is this off by one ?
        self.len() - self.index()
    }
}

impl<WindowIter> Iterator for MelScalingMatrixEnumerator<WindowIter>
    where WindowIter: Iterator<Item=f64>
{
    type Item = (usize, usize, f64);

    #[inline]
    fn next(&mut self) -> Option<(usize, usize, f64)> {
        if self.is_done() {
            return None;
        }

        // start a new row/filter
        if self.is_after_last_col() {
            self.col_index = 0;
            self.row_index += 1;
            if self.is_after_last_row() {
                return None;
            }

            let start_mel = self.start_mels_iter.next()
                .expect("self.start_mels_iter.next()");
            let end_mel = self.end_mels_iter.next()
                .expect("self.end_mels_iter.next()");

            let hertz_from_mel = self.hertz_from_mel;
            let start_hertz = hertz_from_mel(start_mel);
            let end_hertz = hertz_from_mel(end_mel);

            // TODO maybe round or floor or ceil here
            self.window_start = usize_from_f64!((
                start_hertz /
                self.max_hertz *
                f64_from_usize!(self.input_size)).floor());

            let window_end = usize_from_f64!((
                end_hertz /
                self.max_hertz *
                f64_from_usize!(self.input_size)).floor());

            self.window_size = std::cmp::max(
                window_end - self.window_start, 1);

            let window_function = self.window_function;
            self.window_iter = window_function(self.window_size);
        }

        let col = self.col_index;

        let value = if col < self.window_start {
            0.
        } else {
            match self.window_iter.next() {
                Some(value) => {
                    value / f64_from_usize!(self.window_size)
                },
                None => 0.
            }
        };

        self.col_index += 1;
        Some((self.row_index, col, value))
    }

    #[inline]
    fn size_hint(&self) -> (usize, Option<usize>) {
        let remaining = self.remaining();
        (remaining, Some(remaining))
    }
}

impl<WindowIter> ExactSizeIterator for MelScalingMatrixEnumerator<WindowIter>
    where WindowIter: Iterator<Item=f64>
{
    #[inline]
    fn len(&self) -> usize {
        self.input_size * self.output_size
    }
}

/// (or many vectors at once in form of a matrix)
/// to transform that vector into mel space.
/// efficiently
/// rows are mel
/// cols are freq
/// TODO get frequency index pair for nth mel
/// TODO enumerate scaling vector
/// TODO which dimension first
pub fn enumerate_mel_scaling_matrix_base<WindowIter>(
    // TODO more descriptive argument names
    min_hertz: f64,
    max_hertz: f64,
    // TODO shape
    input_size: usize,
    output_size: usize,
    mel_from_hertz: fn(f64) -> f64,
    hertz_from_mel: fn(f64) -> f64,
    window_function: fn(usize) -> WindowIter
) -> MelScalingMatrixEnumerator<WindowIter>
    where WindowIter: Iterator<Item=f64>
{
    assert!(min_hertz < max_hertz);
    assert!(output_size < input_size);
    assert!(0 < output_size);
    assert!(0 < input_size);

    let min_mel = mel_from_hertz(min_hertz);
    let max_mel = mel_from_hertz(max_hertz);

    // initially start_mels_iter == end_mels_iter
    let mut start_mels_iter = linspace(min_mel, max_mel, output_size + 2);
    let mut end_mels_iter = linspace(min_mel, max_mel, output_size + 2);
    end_mels_iter.next();
    end_mels_iter.next();

    let start_mel = start_mels_iter.next().unwrap();
    let end_mel = end_mels_iter.next().unwrap();

    let start_hertz = hertz_from_mel(start_mel);
    let end_hertz = hertz_from_mel(end_mel);

    // TODO maybe round or floor or ceil here
    let start_index = usize_from_f64!((
        start_hertz / max_hertz * f64_from_usize!(input_size)).floor());

    let end_index = usize_from_f64!((
        end_hertz / max_hertz * f64_from_usize!(input_size)).floor());

    let window_size = std::cmp::max(end_index - start_index, 1);

    let window_iter = window_function(window_size);

    MelScalingMatrixEnumerator::<WindowIter> {
        // parameters
        input_size: input_size,
        output_size: output_size,
        window_function: window_function,
        hertz_from_mel: hertz_from_mel,
        mel_from_hertz: mel_from_hertz,
        max_hertz: max_hertz,

        // state
        row_index: 0,
        col_index: 0,

        start_mels_iter: start_mels_iter,
        end_mels_iter: end_mels_iter,

        window_start: start_index,
        window_size: window_size,
        window_iter: window_iter,
    }
}

/// one filter per row
pub fn enumerate_mel_scaling_matrix(
    sample_rate: usize,
    window_size: usize,
    input_size: usize,
    output_size: usize,
) -> MelScalingMatrixEnumerator<apodize::TriangularWindowIter>
{
    enumerate_mel_scaling_matrix_base(
        hertz::rayleigh(
            f64_from_usize!(sample_rate),
            f64_from_usize!(window_size)),
        hertz::nyquist(f64_from_usize!(sample_rate)),
        input_size,
        output_size,
        mel_from_hertz,
        hertz_from_mel,
        apodize::triangular_iter
    )
}