1/**
2 * LPC utility code
3 * Copyright (c) 2006  Justin Ruggles <justin.ruggles@gmail.com>
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22#include "libavutil/lls.h"
23#include "dsputil.h"
24
25#define LPC_USE_DOUBLE
26#include "lpc.h"
27
28
29/**
30 * Apply Welch window function to audio block
31 */
32static void apply_welch_window(const int32_t *data, int len, double *w_data)
33{
34    int i, n2;
35    double w;
36    double c;
37
38    assert(!(len&1)); //the optimization in r11881 does not support odd len
39                      //if someone wants odd len extend the change in r11881
40
41    n2 = (len >> 1);
42    c = 2.0 / (len - 1.0);
43
44    w_data+=n2;
45      data+=n2;
46    for(i=0; i<n2; i++) {
47        w = c - n2 + i;
48        w = 1.0 - (w * w);
49        w_data[-i-1] = data[-i-1] * w;
50        w_data[+i  ] = data[+i  ] * w;
51    }
52}
53
54/**
55 * Calculates autocorrelation data from audio samples
56 * A Welch window function is applied before calculation.
57 */
58void ff_lpc_compute_autocorr(const int32_t *data, int len, int lag,
59                             double *autoc)
60{
61    int i, j;
62    double tmp[len + lag + 1];
63    double *data1= tmp + lag;
64
65    apply_welch_window(data, len, data1);
66
67    for(j=0; j<lag; j++)
68        data1[j-lag]= 0.0;
69    data1[len] = 0.0;
70
71    for(j=0; j<lag; j+=2){
72        double sum0 = 1.0, sum1 = 1.0;
73        for(i=j; i<len; i++){
74            sum0 += data1[i] * data1[i-j];
75            sum1 += data1[i] * data1[i-j-1];
76        }
77        autoc[j  ] = sum0;
78        autoc[j+1] = sum1;
79    }
80
81    if(j==lag){
82        double sum = 1.0;
83        for(i=j-1; i<len; i+=2){
84            sum += data1[i  ] * data1[i-j  ]
85                 + data1[i+1] * data1[i-j+1];
86        }
87        autoc[j] = sum;
88    }
89}
90
91/**
92 * Quantize LPC coefficients
93 */
94static void quantize_lpc_coefs(double *lpc_in, int order, int precision,
95                               int32_t *lpc_out, int *shift, int max_shift, int zero_shift)
96{
97    int i;
98    double cmax, error;
99    int32_t qmax;
100    int sh;
101
102    /* define maximum levels */
103    qmax = (1 << (precision - 1)) - 1;
104
105    /* find maximum coefficient value */
106    cmax = 0.0;
107    for(i=0; i<order; i++) {
108        cmax= FFMAX(cmax, fabs(lpc_in[i]));
109    }
110
111    /* if maximum value quantizes to zero, return all zeros */
112    if(cmax * (1 << max_shift) < 1.0) {
113        *shift = zero_shift;
114        memset(lpc_out, 0, sizeof(int32_t) * order);
115        return;
116    }
117
118    /* calculate level shift which scales max coeff to available bits */
119    sh = max_shift;
120    while((cmax * (1 << sh) > qmax) && (sh > 0)) {
121        sh--;
122    }
123
124    /* since negative shift values are unsupported in decoder, scale down
125       coefficients instead */
126    if(sh == 0 && cmax > qmax) {
127        double scale = ((double)qmax) / cmax;
128        for(i=0; i<order; i++) {
129            lpc_in[i] *= scale;
130        }
131    }
132
133    /* output quantized coefficients and level shift */
134    error=0;
135    for(i=0; i<order; i++) {
136        error -= lpc_in[i] * (1 << sh);
137        lpc_out[i] = av_clip(lrintf(error), -qmax, qmax);
138        error -= lpc_out[i];
139    }
140    *shift = sh;
141}
142
143static int estimate_best_order(double *ref, int min_order, int max_order)
144{
145    int i, est;
146
147    est = min_order;
148    for(i=max_order-1; i>=min_order-1; i--) {
149        if(ref[i] > 0.10) {
150            est = i+1;
151            break;
152        }
153    }
154    return est;
155}
156
157/**
158 * Calculate LPC coefficients for multiple orders
159 *
160 * @param use_lpc LPC method for determining coefficients
161 * 0  = LPC with fixed pre-defined coeffs
162 * 1  = LPC with coeffs determined by Levinson-Durbin recursion
163 * 2+ = LPC with coeffs determined by Cholesky factorization using (use_lpc-1) passes.
164 */
165int ff_lpc_calc_coefs(DSPContext *s,
166                      const int32_t *samples, int blocksize, int min_order,
167                      int max_order, int precision,
168                      int32_t coefs[][MAX_LPC_ORDER], int *shift, int use_lpc,
169                      int omethod, int max_shift, int zero_shift)
170{
171    double autoc[MAX_LPC_ORDER+1];
172    double ref[MAX_LPC_ORDER];
173    double lpc[MAX_LPC_ORDER][MAX_LPC_ORDER];
174    int i, j, pass;
175    int opt_order;
176
177    assert(max_order >= MIN_LPC_ORDER && max_order <= MAX_LPC_ORDER && use_lpc > 0);
178
179    if(use_lpc == 1){
180        s->lpc_compute_autocorr(samples, blocksize, max_order, autoc);
181
182        compute_lpc_coefs(autoc, max_order, &lpc[0][0], MAX_LPC_ORDER, 0, 1);
183
184        for(i=0; i<max_order; i++)
185            ref[i] = fabs(lpc[i][i]);
186    }else{
187        LLSModel m[2];
188        double var[MAX_LPC_ORDER+1], av_uninit(weight);
189
190        for(pass=0; pass<use_lpc-1; pass++){
191            av_init_lls(&m[pass&1], max_order);
192
193            weight=0;
194            for(i=max_order; i<blocksize; i++){
195                for(j=0; j<=max_order; j++)
196                    var[j]= samples[i-j];
197
198                if(pass){
199                    double eval, inv, rinv;
200                    eval= av_evaluate_lls(&m[(pass-1)&1], var+1, max_order-1);
201                    eval= (512>>pass) + fabs(eval - var[0]);
202                    inv = 1/eval;
203                    rinv = sqrt(inv);
204                    for(j=0; j<=max_order; j++)
205                        var[j] *= rinv;
206                    weight += inv;
207                }else
208                    weight++;
209
210                av_update_lls(&m[pass&1], var, 1.0);
211            }
212            av_solve_lls(&m[pass&1], 0.001, 0);
213        }
214
215        for(i=0; i<max_order; i++){
216            for(j=0; j<max_order; j++)
217                lpc[i][j]=-m[(pass-1)&1].coeff[i][j];
218            ref[i]= sqrt(m[(pass-1)&1].variance[i] / weight) * (blocksize - max_order) / 4000;
219        }
220        for(i=max_order-1; i>0; i--)
221            ref[i] = ref[i-1] - ref[i];
222    }
223    opt_order = max_order;
224
225    if(omethod == ORDER_METHOD_EST) {
226        opt_order = estimate_best_order(ref, min_order, max_order);
227        i = opt_order-1;
228        quantize_lpc_coefs(lpc[i], i+1, precision, coefs[i], &shift[i], max_shift, zero_shift);
229    } else {
230        for(i=min_order-1; i<max_order; i++) {
231            quantize_lpc_coefs(lpc[i], i+1, precision, coefs[i], &shift[i], max_shift, zero_shift);
232        }
233    }
234
235    return opt_order;
236}
237