aboutsummaryrefslogtreecommitdiffstats
path: root/libavcodec/lpc.c
blob: 3ed61563ee723cab328d75e7b5143ac418f57860 (plain) (blame)
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
313
314
315
316
317
318
319
320
321
322
323
324
325
326
/*
 * LPC utility code
 * Copyright (c) 2006  Justin Ruggles <justin.ruggles@gmail.com>
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * FFmpeg is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with FFmpeg; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

#include "libavutil/common.h"
#include "libavutil/lls.h"
#include "libavutil/mem_internal.h"

#define LPC_USE_DOUBLE
#include "lpc.h"
#include "libavutil/avassert.h"


/**
 * Apply Welch window function to audio block
 */
static void lpc_apply_welch_window_c(const int32_t *data, int len,
                                     double *w_data)
{
    int i, n2;
    double w;
    double c;

    n2 = (len >> 1);
    c = 2.0 / (len - 1.0);

    if (len & 1) {
        for(i=0; i<n2; i++) {
            w = c - i - 1.0;
            w = 1.0 - (w * w);
            w_data[i] = data[i] * w;
            w_data[len-1-i] = data[len-1-i] * w;
        }
        return;
    }

    w_data+=n2;
      data+=n2;
    for(i=0; i<n2; i++) {
        w = c - n2 + i;
        w = 1.0 - (w * w);
        w_data[-i-1] = data[-i-1] * w;
        w_data[+i  ] = data[+i  ] * w;
    }
}

/**
 * Calculate autocorrelation data from audio samples
 * A Welch window function is applied before calculation.
 */
static void lpc_compute_autocorr_c(const double *data, int len, int lag,
                                   double *autoc)
{
    int i, j;

    for(j=0; j<lag; j+=2){
        double sum0 = 1.0, sum1 = 1.0;
        for(i=j; i<len; i++){
            sum0 += data[i] * data[i-j];
            sum1 += data[i] * data[i-j-1];
        }
        autoc[j  ] = sum0;
        autoc[j+1] = sum1;
    }

    if(j==lag){
        double sum = 1.0;
        for(i=j-1; i<len; i+=2){
            sum += data[i  ] * data[i-j  ]
                 + data[i+1] * data[i-j+1];
        }
        autoc[j] = sum;
    }
}

/**
 * Quantize LPC coefficients
 */
static void quantize_lpc_coefs(double *lpc_in, int order, int precision,
                               int32_t *lpc_out, int *shift, int min_shift,
                               int max_shift, int zero_shift)
{
    int i;
    double cmax, error;
    int32_t qmax;
    int sh;

    /* define maximum levels */
    qmax = (1 << (precision - 1)) - 1;

    /* find maximum coefficient value */
    cmax = 0.0;
    for(i=0; i<order; i++) {
        cmax= FFMAX(cmax, fabs(lpc_in[i]));
    }

    /* if maximum value quantizes to zero, return all zeros */
    if(cmax * (1 << max_shift) < 1.0) {
        *shift = zero_shift;
        memset(lpc_out, 0, sizeof(int32_t) * order);
        return;
    }

    /* calculate level shift which scales max coeff to available bits */
    sh = max_shift;
    while((cmax * (1 << sh) > qmax) && (sh > min_shift)) {
        sh--;
    }

    /* since negative shift values are unsupported in decoder, scale down
       coefficients instead */
    if(sh == 0 && cmax > qmax) {
        double scale = ((double)qmax) / cmax;
        for(i=0; i<order; i++) {
            lpc_in[i] *= scale;
        }
    }

    /* output quantized coefficients and level shift */
    error=0;
    for(i=0; i<order; i++) {
        error -= lpc_in[i] * (1 << sh);
        lpc_out[i] = av_clip(lrintf(error), -qmax, qmax);
        error -= lpc_out[i];
    }
    *shift = sh;
}

static int estimate_best_order(double *ref, int min_order, int max_order)
{
    int i, est;

    est = min_order;
    for(i=max_order-1; i>=min_order-1; i--) {
        if(ref[i] > 0.10) {
            est = i+1;
            break;
        }
    }
    return est;
}

int ff_lpc_calc_ref_coefs(LPCContext *s,
                          const int32_t *samples, int order, double *ref)
{
    double autoc[MAX_LPC_ORDER + 1];

    s->lpc_apply_welch_window(samples, s->blocksize, s->windowed_samples);
    s->lpc_compute_autocorr(s->windowed_samples, s->blocksize, order, autoc);
    compute_ref_coefs(autoc, order, ref, NULL);

    return order;
}

double ff_lpc_calc_ref_coefs_f(LPCContext *s, const float *samples, int len,
                               int order, double *ref)
{
    int i;
    double signal = 0.0f, avg_err = 0.0f;
    double autoc[MAX_LPC_ORDER+1] = {0}, error[MAX_LPC_ORDER+1] = {0};
    const double a = 0.5f, b = 1.0f - a;

    /* Apply windowing */
    for (i = 0; i <= len / 2; i++) {
        double weight = a - b*cos((2*M_PI*i)/(len - 1));
        s->windowed_samples[i] = weight*samples[i];
        s->windowed_samples[len-1-i] = weight*samples[len-1-i];
    }

    s->lpc_compute_autocorr(s->windowed_samples, len, order, autoc);
    signal = autoc[0];
    compute_ref_coefs(autoc, order, ref, error);
    for (i = 0; i < order; i++)
        avg_err = (avg_err + error[i])/2.0f;
    return avg_err ? signal/avg_err : NAN;
}

/**
 * Calculate LPC coefficients for multiple orders
 *
 * @param lpc_type LPC method for determining coefficients,
 *                 see #FFLPCType for details
 */
int ff_lpc_calc_coefs(LPCContext *s,
                      const int32_t *samples, int blocksize, int min_order,
                      int max_order, int precision,
                      int32_t coefs[][MAX_LPC_ORDER], int *shift,
                      enum FFLPCType lpc_type, int lpc_passes,
                      int omethod, int min_shift, int max_shift, int zero_shift)
{
    double autoc[MAX_LPC_ORDER+1];
    double ref[MAX_LPC_ORDER] = { 0 };
    double lpc[MAX_LPC_ORDER][MAX_LPC_ORDER];
    int i, j, pass = 0;
    int opt_order;

    av_assert2(max_order >= MIN_LPC_ORDER && max_order <= MAX_LPC_ORDER &&
           lpc_type > FF_LPC_TYPE_FIXED);
    av_assert0(lpc_type == FF_LPC_TYPE_CHOLESKY || lpc_type == FF_LPC_TYPE_LEVINSON);

    /* reinit LPC context if parameters have changed */
    if (blocksize != s->blocksize || max_order != s->max_order ||
        lpc_type  != s->lpc_type) {
        ff_lpc_end(s);
        ff_lpc_init(s, blocksize, max_order, lpc_type);
    }

    if(lpc_passes <= 0)
        lpc_passes = 2;

    if (lpc_type == FF_LPC_TYPE_LEVINSON || (lpc_type == FF_LPC_TYPE_CHOLESKY && lpc_passes > 1)) {
        s->lpc_apply_welch_window(samples, blocksize, s->windowed_samples);

        s->lpc_compute_autocorr(s->windowed_samples, blocksize, max_order, autoc);

        compute_lpc_coefs(autoc, max_order, &lpc[0][0], MAX_LPC_ORDER, 0, 1);

        for(i=0; i<max_order; i++)
            ref[i] = fabs(lpc[i][i]);

        pass++;
    }

    if (lpc_type == FF_LPC_TYPE_CHOLESKY) {
        LLSModel *m = s->lls_models;
        LOCAL_ALIGNED(32, double, var, [FFALIGN(MAX_LPC_ORDER+1,4)]);
        double av_uninit(weight);
        memset(var, 0, FFALIGN(MAX_LPC_ORDER+1,4)*sizeof(*var));

        for(j=0; j<max_order; j++)
            m[0].coeff[max_order-1][j] = -lpc[max_order-1][j];

        for(; pass<lpc_passes; pass++){
            avpriv_init_lls(&m[pass&1], max_order);

            weight=0;
            for(i=max_order; i<blocksize; i++){
                for(j=0; j<=max_order; j++)
                    var[j]= samples[i-j];

                if(pass){
                    double eval, inv, rinv;
                    eval= m[pass&1].evaluate_lls(&m[(pass-1)&1], var+1, max_order-1);
                    eval= (512>>pass) + fabs(eval - var[0]);
                    inv = 1/eval;
                    rinv = sqrt(inv);
                    for(j=0; j<=max_order; j++)
                        var[j] *= rinv;
                    weight += inv;
                }else
                    weight++;

                m[pass&1].update_lls(&m[pass&1], var);
            }
            avpriv_solve_lls(&m[pass&1], 0.001, 0);
        }

        for(i=0; i<max_order; i++){
            for(j=0; j<max_order; j++)
                lpc[i][j]=-m[(pass-1)&1].coeff[i][j];
            ref[i]= sqrt(m[(pass-1)&1].variance[i] / weight) * (blocksize - max_order) / 4000;
        }
        for(i=max_order-1; i>0; i--)
            ref[i] = ref[i-1] - ref[i];
    }

    opt_order = max_order;

    if(omethod == ORDER_METHOD_EST) {
        opt_order = estimate_best_order(ref, min_order, max_order);
        i = opt_order-1;
        quantize_lpc_coefs(lpc[i], i+1, precision, coefs[i], &shift[i],
                           min_shift, max_shift, zero_shift);
    } else {
        for(i=min_order-1; i<max_order; i++) {
            quantize_lpc_coefs(lpc[i], i+1, precision, coefs[i], &shift[i],
                               min_shift, max_shift, zero_shift);
        }
    }

    return opt_order;
}

av_cold int ff_lpc_init(LPCContext *s, int blocksize, int max_order,
                        enum FFLPCType lpc_type)
{
    s->blocksize = blocksize;
    s->max_order = max_order;
    s->lpc_type  = lpc_type;

    s->windowed_buffer = av_mallocz((blocksize + 2 + FFALIGN(max_order, 4)) *
                                    sizeof(*s->windowed_samples));
    if (!s->windowed_buffer)
        return AVERROR(ENOMEM);
    s->windowed_samples = s->windowed_buffer + FFALIGN(max_order, 4);

    s->lpc_apply_welch_window = lpc_apply_welch_window_c;
    s->lpc_compute_autocorr   = lpc_compute_autocorr_c;

    if (ARCH_X86)
        ff_lpc_init_x86(s);

    return 0;
}

av_cold void ff_lpc_end(LPCContext *s)
{
    av_freep(&s->windowed_buffer);
}