History

spectrogram effect

robs committed 18 years ago

Commit 52f4fa01 parent 7cacd5310 files · +677 −7Browse files

Unified · Split

10 files
  1. AUTHORS +2 −2
  2. CMakeLists.txt +1 −0
  3. ChangeLog +1 −0
  4. README +4 −4
  5. configure.ac +21 −0
  6. soxeffect.7 +104 −0
  7. src/Makefile.am +4 −1
  8. src/effects.h +3 −0
  9. src/soxconfig.h.cmake +1 −0
  10. src/spectrogram.c +536 −0
AUTHORS+2−2
@@ -9,7 +9,7 @@ Creator: · 8 unmodified lines
99 Mantainers:
1010 Chris Bagwell cbagwell@users.sourceforge.net
1111 OSS and Sun players, bugfixes, ADPCM support,
12− patch collection and maintance.
12+ patch collection and maintenance.
1313 Reuben Thomas rrt@sc3d.org
1414 Build system makeover, libsndfile and ffmpeg sound
1515 format support, libao playback, Secret Rabbit Code
@@ -21,7 +21,7 @@ Mantainers: · 5 unmodified lines
2121 EFFECTS: key, tempo, pad, bass, treble, delay, new reverb, new
2222 flanger, soft-knee companding, speed via resampling, filters
2323 makeover inc. gnuplot & octave plotting, splice, remix, norm,
24− contrast, new rate;
24+ contrast, new rate, spectrogram;
2525 new effects chain with buffering and any # channels.
2626 OTHERS: open input files via URL, file merge, play
2727 multiple files with mixed format types, play with replay-gain,
CMakeLists.txt+1−0
@@ -73,6 +73,7 @@ optional(HAVE_LAME_LAME_H lame/lame.h mp3lame lame_init mp3) · 72 unmodified lines
7373 if (NOT HAVE_LAME_LAME_H)
7474 optional(HAVE_LAME_LAME_H lame.h mp3lame lame_init mp3)
7575 endif (NOT HAVE_LAME_LAME_H)
76+optional(HAVE_PNG png.h png png_set_rows spectrogram)
7677 optional(HAVE_SAMPLERATE_H samplerate.h samplerate src_simple rabbit)
7778 optional(HAVE_SNDFILE_H sndfile.h sndfile sf_open sndfile)
7879 optional(HAVE_SNDFILE_H sndfile.h sndfile sf_open fap)
ChangeLog+1−0
@@ -57,6 +57,7 @@ Effects: · 56 unmodified lines
5757 o New `delay' effect; delay one or more channels. (robs)
5858 o New `contrast' enhancement effect [FR 708923]. (robs)
5959 o Rewrite of rate; resample, polyphase, & rabbit now deprecated. (robs)
60+ o New `spectrogram' effect; creates a PNG. (robs)
6061 o `synth' can now sweep linearly (as well as logarithmically). (robs)
6162 o Fix synth max. level setting for some output encodings. (robs)
6263 o Fix crash on 64-bit arch. with tempo & key effects. (Sami Liedes)
README+4−4
@@ -71,9 +71,11 @@ The sound effects include: · 70 unmodified lines
7171 o Band-reject filter
7272 o Bass tone control
7373 o Compress and Expand (compand) the dynamic range of a sample
74+ o Contrast enhancement effect
7475 o Chorus effect
7576 o DC-shift audio - useful to get the best volume adjustments
7677 o De-emphasis filter
78+ o Delay one or more channels
7779 o Move sound stage of CD audio to in front of you (for headphone
7880 use)
7981 o Add an echo
@@ -91,15 +93,13 @@ The sound effects include: · 11 unmodified lines
9193 o Apply a phaser effect
9294 o Change the key (pitch) of a sound file without affecting its speed
9395 o Repeat audio data
94− o Change sampling rates using several different algorithms. A
95− 'resample' and 'polyphase' effect use high-grade signal rate
96− changes using real signal theory!
96+ o Change sampling rate
9797 o Apply a reverb effect
9898 o Reverse the sound samples (to search for Satanic messages ;-)
9999 o Detect periods of silence and start and stop processing based on
100100 it
101+ o Create a spectrogram PNG
101102 o Change speed (pitch and tempo together)
102− motor on a tape recorder)
103103 o Splice together audio sections
104104 o Display general stats on a sound sample
105105 o Change the tempo an audio file (without affecting pitch)
configure.ac+21−0
@@ -414,6 +414,26 @@ fi · 413 unmodified lines
414414 AM_CONDITIONAL(HAVE_WAVPACK, test x$using_wavpack = xyes)
415415 AC_SUBST(WAVPACK_LIBS)
416416
417+dnl Check for png libraries
418+AC_ARG_WITH(png,
419+ AC_HELP_STRING([--without-png],
420+ [Don't try to use png]))
421+using_png=no
422+if test "$with_png" != "no"; then
423+ using_png=yes
424+ AC_CHECK_HEADER(png.h,
425+ [AC_CHECK_LIB(png, png_set_rows, PNG_LIBS="$PNG_LIBS -lpng -lz" ,using_png=no)],
426+ using_png=no)
427+ if test "$with_png" = "yes" -a "$using_png" = "no"; then
428+ AC_MSG_FAILURE([cannot find png])
429+ fi
430+fi
431+if test "$using_png" = yes; then
432+ AC_DEFINE(HAVE_PNG, 1, [Define to 1 if you have PNG.])
433+fi
434+AM_CONDITIONAL(HAVE_PNG, test x$using_png = xyes)
435+AC_SUBST(PNG_LIBS)
436+
417437 dnl Test for libsamplerate.
418438 AC_ARG_WITH(samplerate,
419439 AC_HELP_STRING([--without-samplerate],
@@ -483,6 +503,7 @@ echo "AMR-WB format..................... $using_amr_wb" · 63 unmodified lines
483503 echo "AMR-NB format..................... $using_amr_nb"
484504 echo "WavPack format.................... $using_wavpack"
485505 echo "LADSPA effects.................... $using_ladspa"
506+echo "PNG support....................... $using_png"
486507 echo "Secret Rabbit Code resampling..... $using_samplerate"
487508 echo "pkg-config location............... $pkgconfig_option"
488509 echo
soxeffect.7+104−0
@@ -1027,6 +1027,110 @@ effect using its default quality/speed. For higher quality or higher speed · 1026 unmodified lines
10271027 resampling, in addition to the \fBspeed\fR effect, specify
10281028 the \fBrate\fR effect with the desired quality options.
10291029 .TP
1030+\fBspectrogram \fR[options]
1031+Create a spectrogram of the audio. This effect is optional; type \fBsox
1032+\-\-help\fR and check the list of supported effects to see if it has
1033+been included.
1034+.SP
1035+The spectrogram is rendered in a Portable Network Graphic (PNG) file,
1036+and shows time in the X-axis, frequency in the Y-axis, and audio signal
1037+magnitude in the Z-axis. Z-axis values are represented by the colour
1038+(or intensity) of the pixels in the X-Y plane.
1039+.RS
1040+.IP \fB-x\ \fInum\fR
1041+X-axis pixels/second, default 100. This controls the width of the
1042+spectrogram;
1043+.I num
1044+can be from 1 (low time resolution) to 5000 (high time resolution). SoX
1045+may make a slight adjustment to the given number for processing
1046+quantisation reasons; if so, SoX will report the actual number used
1047+(viewable when
1048+.B \-\-verbose
1049+is in effect).
1050+.SP
1051+The maximum width of the spectrogram is 999 pixels; if the audio length
1052+and the given
1053+.B \-x
1054+number are such that this would be exceeded, then the spectrogram (and
1055+the effects chain) will be truncated. To move the spectrogram to a
1056+point later in the audio stream, first invoke the
1057+.B trim
1058+effect; e.g.
1059+.EX
1060+ sox audio.ogg -n trim 1:00 spectrogram
1061+.EE
1062+starts the spectrogram at 1 minute through the audio.
1063+.IP \fB-y\ \fInum\fR
1064+Y-axis resolution (1 - 4), default 2.
1065+This controls the height of the spectrogram;
1066+.I num
1067+can be from 1 (low frequency resolution) to 4 (high frequency
1068+resolution). For values greater than 2, the resulting image may be too
1069+tall to display on the screen; if so, a graphic manipulation package
1070+(such as
1071+.BR
1072+ImageMagick (1))
1073+can be used to re-size the image.
1074+.SP
1075+To increase the frequency resolution without increasing the height of
1076+the spectrogram, the
1077+.B rate
1078+effect may be invoked to reduce the sampling rate of the signal before
1079+invoking
1080+.BR spectrogram ;
1081+e.g.
1082+.EX
1083+ sox audio.ogg -r 4k -n rate spectrogram
1084+.EE
1085+allows detailed analysis of frequencies up to 2kHz (half the sampling
1086+rate).
1087+.IP \fB-z\ \fInum\fR
1088+Z-axis (colour) range in dB, default 120. This sets the dynamic-range
1089+of the spectrogram to be \-\fInum\fR\ dBFS to 0\ dBFS.
1090+.I Num
1091+may range from 40 to 180. Decreasing dynamic-range, effectively
1092+increases the `contrast' of the spectrogram display, and vice versa.
1093+.IP \fB-g\ \fInum\fR
1094+Apply
1095+.I num
1096+dB gain (or attenuation if
1097+.I num
1098+is negative) to the audio before displaying it. Appling gain or
1099+attenuation shifts the the range of the Z-axis and effectively controls
1100+the `brightness' of the spectrogram display. Note that the audio in the
1101+SoX effects chain remains unaffected by this option (cf. the
1102+.B gain
1103+effect).
1104+.IP \fB-w\ \fIname\fR
1105+Window: Hann (default), Hamming, Bartlett, Rectangular, Kaiser. The
1106+spectrogram is produced using the Discrete Fourier Transform (DFT)
1107+algorithm. A significant paramater to this algorithm is the choice of
1108+`window function'. By default, SoX uses the Hann window which has good
1109+all-round frequency-resolution and dynamic-range properties. For better
1110+frequency resolution (but lower dynamic-range), select a Hamming window;
1111+for higher dynamic-range (but poorer frequency-resolution), select a
1112+Kaiser window. Selecting a window other than Hann will usually require
1113+a corresponding
1114+.B \-z
1115+setting.
1116+.IP \fB-s\ \fInum\fR
1117+Display style (1 - 4), default 1. There are 2 colour and 2 monochrome
1118+styles; select the style that works best with the given audio signal and
1119+your display or printer.
1120+.IP \fB-a\fR
1121+Toggle the presence of the axes lines. This is sometimes useful in
1122+helping to discern artefacts at the spectrogram edges.
1123+.IP \fB-t\ \fItext\fR
1124+Text to display above the spectrogram.
1125+.IP \fB-c\ \fItext\fR
1126+Text to display below and to the left of the spectrogram.
1127+.IP \fB-o\ \fItext\fR
1128+Name of the spectrogram output PNG file, default `spectrogram.png'.
1129+.RE
1130+For the ability to perform off-line processing of spectral data, see the
1131+.B stat
1132+effect.
1133+.TP
10301134 \fBsplice \fR { \fIposition\fR[\fB,\fIexcess\fR[\fB,\fIleeway\fR]] }
10311135 Splice together audio sections. This effect provides two things over
10321136 simple audio concatenation: a (usually short) cross-fade is applied at
src/Makefile.am+4−1
@@ -231,9 +231,12 @@ libsfx_la_SOURCES = \ · 230 unmodified lines
231231 remix.c repeat.c resample.c reverb.c reverse.c silence.c skeleff.c \
232232 speed.c splice.c stat.c stretch.c swap.c synth.c tempo.c tremolo.c \
233233 trim.c vol.c
234+if HAVE_PNG
235+ libsfx_la_SOURCES += spectrogram.c
236+endif
234237
235238 libsfx_la_CFLAGS = @SAMPLERATE_CFLAGS@
236−libsfx_la_LIBADD = @SAMPLERATE_LIBS@ libsox.la
239+libsfx_la_LIBADD = @PNG_LIBS@ @SAMPLERATE_LIBS@ libsox.la
237240
238241 ############################################
239242 # libsox - file format and utility library #
src/effects.h+3−0
@@ -64,6 +64,9 @@ · 63 unmodified lines
6464 EFFECT(reverb)
6565 EFFECT(reverse)
6666 EFFECT(silence)
67+#ifdef HAVE_PNG
68+ EFFECT(spectrogram)
69+#endif
6770 EFFECT(speed)
6871 EFFECT(splice)
6972 EFFECT(stat)
src/soxconfig.h.cmake+1−0
@@ -19,6 +19,7 @@ · 18 unmodified lines
1919 #cmakedefine HAVE_MAD_H 1
2020 #cmakedefine HAVE_OGG_VORBIS 1
2121 #cmakedefine HAVE_POPEN 1
22+#cmakedefine HAVE_PNG 1
2223 #cmakedefine HAVE_SAMPLERATE_H 1
2324 #cmakedefine HAVE_SNDFILE_1_0_12 1
2425 #cmakedefine HAVE_SNDFILE_H 1
src/spectrogram.c+536−0
@@ -0,0 +1,536 @@
1+/* libSoX effect: Spectrogram (c) 2008 robs@users.sourceforge.net
2+ *
3+ * This library is free software; you can redistribute it and/or modify it
4+ * under the terms of the GNU Lesser General Public License as published by
5+ * the Free Software Foundation; either version 2.1 of the License, or (at
6+ * your option) any later version.
7+ *
8+ * This library is distributed in the hope that it will be useful, but
9+ * WITHOUT ANY WARRANTY; without even the implied warranty of
10+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser
11+ * General Public License for more details.
12+ *
13+ * You should have received a copy of the GNU Lesser General Public License
14+ * along with this library; if not, write to the Free Software Foundation,
15+ * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
16+ */
17+
18+#ifdef NDEBUG /* Enable assert always. */
19+#undef NDEBUG /* Must undef above assert.h or other that might include it. */
20+#endif
21+
22+#include "sox_i.h"
23+#include "fft4g.h"
24+#include "getopt.h"
25+#include <assert.h>
26+#include <math.h>
27+#include <png.h>
28+
29+#define malloc lsx_malloc
30+#define calloc lsx_calloc
31+#define realloc lsx_realloc
32+#define FROM_SOX SOX_SAMPLE_TO_FLOAT_64BIT
33+#define DFT_BASE_SIZE 512
34+#define MAX_DFT_SIZE_SHIFT 3
35+#define MAX_DFT_SIZE (DFT_BASE_SIZE << MAX_DFT_SIZE_SHIFT)
36+#define MAX_COLS 999 /* Also max seconds */
37+
38+static void apply_hann(double h[], const int num_points)
39+{
40+ int i, m = num_points - 1;
41+ for (i = 0; i < num_points; ++i) {
42+ double x = 2 * M_PI * i / m;
43+ h[i] *= .5 - .5 * cos(x);
44+ }
45+}
46+
47+static void apply_hamming(double h[], const int num_points)
48+{
49+ int i, m = num_points - 1;
50+ for (i = 0; i < num_points; ++i) {
51+ double x = 2 * M_PI * i / m;
52+ h[i] *= .53836 - .46164 * cos(x);
53+ }
54+}
55+
56+static void apply_bartlett(double h[], const int num_points)
57+{
58+ int i, m = num_points - 1;
59+ for (i = 0; i < num_points; ++i) {
60+ h[i] *= 2. / m * (m / 2. - fabs(i - m / 2.));
61+ }
62+}
63+
64+static double kaiser_beta(double att)
65+{
66+ if (att > 100 ) return .1117 * att - 1.11;
67+ if (att > 50 ) return .1102 * (att - 8.7);
68+ if (att > 20.96) return .58417 * pow(att -20.96, .4) + .07886 * (att - 20.96);
69+ return 0;
70+}
71+
72+static void apply_kaiser(double h[], const int num_points, double beta)
73+{
74+ int i, m = num_points - 1;
75+ for (i = 0; i <= m; ++i) {
76+ double x = 2. * i / m - 1;
77+ h[i] *= bessel_I_0(beta * sqrt(1 - x * x)) / bessel_I_0(beta);
78+ }
79+}
80+
81+typedef enum {Window_Hann, Window_Hamming, Window_Bartlett, Window_Rectangular, Window_Kaiser} win_type_t;
82+static enum_item const window_options[] = {
83+ ENUM_ITEM(Window_,Hann)
84+ ENUM_ITEM(Window_,Hamming)
85+ ENUM_ITEM(Window_,Bartlett)
86+ ENUM_ITEM(Window_,Rectangular)
87+ ENUM_ITEM(Window_,Kaiser)
88+ {0, 0}};
89+
90+typedef struct {
91+ int y_size, dft_size, rows, cols, step_size, block_steps, block_num;
92+ double buf[MAX_DFT_SIZE];
93+ double dft_buf[MAX_DFT_SIZE];
94+ double window[MAX_DFT_SIZE];
95+ int bit_rev_table[100]; /* For Ooura fft */
96+ double sin_cos_table[dft_sc_len(MAX_DFT_SIZE)]; /* ditto */
97+ double magnitudes[(MAX_DFT_SIZE >> 1) + 1];
98+ float * dBfs;
99+ double pixels_per_sec, block_norm, max;
100+ int dB_range, gain, style, read, end, end_min, last_end;
101+ char const * out_name, * title, * comment;
102+ sox_bool no_axes;
103+ win_type_t win_type;
104+} priv_t;
105+
106+#define GETOPT_NUMERIC(ch, name, min, max) case ch:{ \
107+ char * end_ptr; \
108+ double d = strtod(optarg, &end_ptr); \
109+ if (end_ptr == optarg || d < min || d > max || *end_ptr != '\0') {\
110+ sox_fail("parameter `%s' must be between %g and %g", #name, (double)min, (double)max); \
111+ return lsx_usage(effp); \
112+ } \
113+ p->name = d; \
114+ break; \
115+}
116+
117+static int enum_option(int c, enum_item const * items)
118+{
119+ enum_item const * p = find_enum_text(optarg, items);
120+ if (p == NULL) {
121+ unsigned len = 1;
122+ char * set = lsx_malloc(len);
123+ *set = 0;
124+ for (p = items; p->text; ++p) {
125+ set = lsx_realloc(set, len += 2 + strlen(p->text));
126+ strcat(set, ", "); strcat(set, p->text);
127+ }
128+ sox_fail("-%c: '%s' is not one of: %s.", c, optarg, set + 2);
129+ free(set);
130+ return INT_MAX;
131+ }
132+ return p->value;
133+}
134+
135+static int getopts(sox_effect_t * effp, int argc, char **argv)
136+{
137+ priv_t * p = (priv_t *)effp->priv;
138+ int c, callers_optind = optind, callers_opterr = opterr;
139+
140+ assert(array_length(p->bit_rev_table) >= (size_t)dft_br_len(MAX_DFT_SIZE));
141+
142+ --argv, ++argc, optind = 1, opterr = 0; /* re-jig for getopt */
143+ p->pixels_per_sec = 100, p->y_size = 2, p->dB_range = 120;/* non-0 defaults */
144+ p->style = 1, p->out_name = "spectrogram.png", p->comment = "Created by SoX";
145+
146+ while ((c = getopt(argc, argv, "+x:y:z:g:s:w:t:c:o:a")) != -1) switch (c) {
147+ GETOPT_NUMERIC('x', pixels_per_sec, 1 , 5000)
148+ GETOPT_NUMERIC('y', y_size , 1 , 1 + MAX_DFT_SIZE_SHIFT)
149+ GETOPT_NUMERIC('z', dB_range , 40 , 180)
150+ GETOPT_NUMERIC('g', gain ,-100, 100)
151+ GETOPT_NUMERIC('s', style , 1 , 4)
152+ case 'w': p->win_type = enum_option(c, window_options); break;
153+ case 't': p->title = optarg; break;
154+ case 'c': p->comment = optarg; break;
155+ case 'o': p->out_name = optarg; break;
156+ case 'a': p->no_axes = sox_true; break;
157+ default: sox_fail("unknown option `-%c'", optopt); return lsx_usage(effp);
158+ }
159+ p->y_size -= 1, p->style -= 1;
160+ argc -= optind, optind = callers_optind, opterr = callers_opterr;
161+ return argc || p->win_type == INT_MAX? lsx_usage(effp) : SOX_SUCCESS;
162+}
163+
164+static int make_window(priv_t * p, int end)
165+{
166+ double sum = 0, * w = end < 0? p->window : p->window + end;
167+ int i, n = p->dft_size - abs(end);
168+
169+ if (end) memset(p->window, 0, sizeof(p->window));
170+ for (i = 0; i < n; ++i) w[i] = 1;
171+ switch (p->win_type) {
172+ case Window_Hann: apply_hann(w, n); break;
173+ case Window_Hamming: apply_hamming(w, n); break;
174+ case Window_Bartlett: apply_bartlett(w, n); break;
175+ case Window_Rectangular: break;
176+ default: apply_kaiser(w, n, kaiser_beta(p->dB_range + 10.));
177+ }
178+ for (i = 0; i < p->dft_size; ++i) sum += p->window[i];
179+ for (i = 0; i < p->dft_size; ++i) p->window[i] *= 2 / sum
180+ * sqr((double)n / p->dft_size); /* empirical small window adjustment */
181+ return sum + .5;
182+}
183+
184+static int start(sox_effect_t * effp)
185+{
186+ priv_t * p = (priv_t *)effp->priv;
187+ double actual;
188+
189+ if (effp->in_signal.channels != 1) {
190+ sox_fail("only 1 channel is supported");
191+ return SOX_EOF;
192+ }
193+ p->end = p->dft_size = DFT_BASE_SIZE << p->y_size;
194+ p->rows = (p->dft_size >> 1) + 1;
195+ p->step_size = make_window(p, p->last_end = 0);
196+ p->block_steps = effp->in_signal.rate / p->pixels_per_sec;
197+ p->step_size = p->block_steps / ceil((double)p->block_steps / p->step_size) +.5;
198+ p->block_steps = floor((double)p->block_steps / p->step_size +.5);
199+ p->block_norm = 1. / p->block_steps;
200+ actual = effp->in_signal.rate / p->step_size / p->block_steps;
201+ if (actual != p->pixels_per_sec)
202+ sox_report("actual pixels/s = %g", actual);
203+ sox_debug("step_size=%i block_steps=%i", p->step_size, p->block_steps);
204+ p->max = -p->dB_range;
205+ p->read = (p->step_size - p->dft_size) / 2;
206+ return SOX_SUCCESS;
207+}
208+
209+static int do_column(sox_effect_t * effp)
210+{
211+ priv_t * p = (priv_t *)effp->priv;
212+ int i;
213+
214+ if (p->cols == MAX_COLS) {
215+ sox_warn("PNG truncated at %g seconds", (double)MAX_COLS * p->step_size * p->block_steps / effp->in_signal.rate);
216+ return SOX_EOF;
217+ }
218+ ++p->cols;
219+ p->dBfs = realloc(p->dBfs, p->cols * p->rows * sizeof(*p->dBfs));
220+ for (i = 0; i < p->rows; ++i) {
221+ double dBfs = 10 * log10(p->magnitudes[i] * p->block_norm);
222+ p->dBfs[(p->cols - 1) * p->rows + i] = dBfs + p->gain;
223+ if (dBfs > p->max)
224+ p->max = dBfs;
225+ }
226+ memset(p->magnitudes, 0, p->rows * sizeof(*p->magnitudes));
227+ p->block_num = 0;
228+ return SOX_SUCCESS;
229+}
230+
231+static int flow(sox_effect_t * effp,
232+ const sox_sample_t * ibuf, sox_sample_t * obuf,
233+ sox_size_t * isamp, sox_size_t * osamp)
234+{
235+ priv_t * p = (priv_t *)effp->priv;
236+ sox_size_t len = min(*isamp, *osamp), dummy = 0; /* No need to clip count */
237+ int i;
238+
239+ memcpy(obuf, ibuf, len * sizeof(*obuf)); /* Pass on audio unaffected */
240+ *isamp = *osamp = len;
241+
242+ while (sox_true) {
243+ if (p->read == p->step_size) {
244+ memmove(p->buf, p->buf + p->step_size,
245+ (p->dft_size - p->step_size) * sizeof(*p->buf));
246+ p->read = 0;
247+ }
248+ for (; len && p->read < p->step_size; --len, ++p->read, --p->end)
249+ p->buf[p->dft_size - p->step_size + p->read] = FROM_SOX(*ibuf++, dummy);
250+ if (p->read != p->step_size)
251+ break;
252+
253+ if ((p->end = max(p->end, p->end_min)) != p->last_end)
254+ make_window(p, p->last_end = p->end);
255+ for (i = 0; i < p->dft_size; ++i) p->dft_buf[i] = p->buf[i] * p->window[i];
256+ rdft(p->dft_size, 1, p->dft_buf, p->bit_rev_table, p->sin_cos_table);
257+ p->magnitudes[0] += sqr(p->dft_buf[0]);
258+ for (i = 1; i < p->dft_size >> 1; ++i)
259+ p->magnitudes[i] += sqr(p->dft_buf[2*i]) + sqr(p->dft_buf[2*i+1]);
260+ p->magnitudes[p->dft_size >> 1] += sqr(p->dft_buf[1]);
261+
262+ if (++p->block_num == p->block_steps && do_column(effp) == SOX_EOF)
263+ return SOX_EOF;
264+ }
265+ return SOX_SUCCESS;
266+}
267+
268+static int drain(sox_effect_t * effp, sox_sample_t * obuf_, sox_size_t * osamp)
269+{
270+ priv_t * p = (priv_t *)effp->priv;
271+
272+ if (p->cols != MAX_COLS) {
273+ sox_sample_t * ibuf = calloc(p->dft_size, sizeof(*ibuf));
274+ sox_sample_t * obuf = calloc(p->dft_size, sizeof(*obuf));
275+ sox_size_t isamp = (p->dft_size - p->step_size) / 2;
276+ int left_over = (isamp + p->read) % p->step_size;
277+
278+ if (left_over >= p->step_size >> 1)
279+ isamp += p->step_size - left_over;
280+ sox_debug("cols=%i left=%i end=%i", p->cols, p->read, p->end);
281+ p->end = 0;
282+ p->end_min = -p->dft_size;
283+ if (flow(effp, ibuf, obuf, &isamp, &isamp) == SOX_SUCCESS && p->block_num) {
284+ p->block_norm *= (double)p->block_steps / p->block_num;
285+ do_column(effp);
286+ }
287+ sox_debug("flushed cols=%i left=%i end=%i", p->cols, p->read, p->end);
288+ free(obuf);
289+ free(ibuf);
290+ }
291+ (void)obuf_, *osamp = 0;
292+ return SOX_SUCCESS;
293+}
294+
295+#define spectrum_points (1 + 250)
296+enum {Background = spectrum_points, Text, Labels, Grid, palette_size};
297+
298+static unsigned colour(double x, int range)
299+{
300+ x = range_limit(x, -range, 0);
301+ return (1 + x / range) * (spectrum_points - 1) + .5;
302+}
303+
304+static void make_palette(png_byte * palette, int style)
305+{
306+ int i, j, at;
307+ for (i = j = 0; j < spectrum_points; i += 3, ++j) {
308+ double r, g, b, x = (double)j / (spectrum_points - 1);
309+ if (style > 1)
310+ palette[i+2] = palette[i+1] = palette[i+0] = style == 2? j : 255 - j;
311+ else {
312+ at = style? (spectrum_points - 1) * 3 - i : i;
313+ if (x < .16) r = 0;
314+ else if (x < .72) r = .99 * sin((x - .16) / .56 * M_PI / 2);
315+ else r = .01 * (x - .72) / .28 + .99;
316+ if (x < .52) g = 0;
317+ else if (x < .86) g = .99 * .5 * (1 - cos((x - .52) / .34 * M_PI));
318+ else g = .01 * (x - .86) / .14 + .99;
319+ if (x < .34) b = .5 * sin((x - .00) / .34 * M_PI / 2);
320+ else if (x < .63) b = .5 * .5 * (1 + cos((x - .34) / .29 * M_PI));
321+ else b = (1 - cos((x - .63) / .37 * M_PI / 2));
322+ palette[at + 0] = r * 255 + .5;
323+ palette[at + 1] = g * 255 + .5;
324+ palette[at + 2] = b * 255 + .5;
325+ }
326+ }
327+ i = spectrum_points * 3;
328+ switch (style) {
329+ case 0: case 2:
330+ memcpy(palette + i, "\0\0\0" , 3), i+= 3;
331+ memcpy(palette + i, "\377\377\377", 3), i+= 3;
332+ memcpy(palette + i, "\277\277\277", 3), i+= 3;
333+ memcpy(palette + i, "\177\177\177", 3), i+= 3;
334+ return;
335+ case 1: memcpy(palette + i, "\335\330\320", 3), i+= 3; break;
336+ default: memcpy(palette + i, "\337\337\337", 3), i+= 3;
337+ }
338+ memcpy(palette + i, "\0\0\0" , 3), i+= 3;
339+ memcpy(palette + i, "\077\077\077", 3), i+= 3;
340+ memcpy(palette + i, "\077\077\077", 3), i+= 3;
341+}
342+
343+static const Bytef fixed[] =
344+ "x\332eT\241\266\2450\fDVV>Y\371$re%2\237\200|2\22YY\211D\"+\337'<y\345\312"
345+ "\375\fd\345f\222\224\313\236\235{\270\344L\247a\232\4\246\351\201d\230\222"
346+ "\304D\364^ \352\362S\"m\347\311\237\237\27\64K\243\2302\265\35\v\371<\363y"
347+ "\354_\226g\354\214)e \2458\341\17\20J4\215[z<\271\277\367\0\63\64@\177\330c"
348+ "\227\204 Ir.\5$U\200\260\224\326S\17\200=\\k\20QA\334%\342\20*\303P\234\211"
349+ "\366\36#\370R\276_\316s-\345\222Dlz\363my*;\217\373\346z\267\343\236\364\246"
350+ "\236\365\2419\305p\333\267\23(\207\265\333\233\325Y\342\243\265\357\262\215"
351+ "\263t\271$\276\226ea\271.\367&\320\347\202_\234\27\377\345\222\253?\3422\364"
352+ "\207y\256\236\229\331\33\f\376\227\266\"\356\253j\366\363\347\334US\34]\371?"
353+ "\255\371\336\372z\265v\34\226\247\32\324\217\334\337\317U4\16\316{N\370\31"
354+ "\365\357iL\231y\33y\264\211D7\337\4\244\261\220D\346\1\261\357\355>\3\342"
355+ "\223\363\0\303\277\f[\214A,p\34`\255\355\364\37\372\224\342\277\f\207\255\36"
356+ "_V\7\34\241^\316W\257\177\b\242\300\34\f\276\33?/9_\331f\346\36\25Y)\2301"
357+ "\257\2414|\35\365\237\3424k\3\244\3\242\261\6\b\275>z$\370\215:\270\363w\36/"
358+ "\265kF\v\20o6\242\301\364\336\27\325\257\321\364fs\231\215G\32=\257\305Di"
359+ "\304^\177\304R\364Q=\225\373\33\320\375\375\372\200\337\37\374}\334\337\20"
360+ "\364\310(]\304\267b\177\326Yrj\312\277\373\233\37\340";
361+static unsigned char * font;
362+#define font_x 5
363+#define font_y 12
364+#define font_X (font_x + 1)
365+
366+#define pixel(x,y) pixels[(y) * cols + (x)]
367+#define print_at(x,y,c,t) print_at_(pixels,cols,x,y,c,t,0)
368+#define print_up(x,y,c,t) print_at_(pixels,cols,x,y,c,t,1)
369+
370+static void print_at_(png_byte * pixels, int cols, int x, int y, int c, char const * text, int orientation)
371+{
372+ for (;*text; ++text) {
373+ int pos = (*text - ' ') * font_y;
374+ int i, j;
375+ for (i = 0; i < font_y; ++i) {
376+ unsigned line = font[pos++];
377+ for (j = 0; j < font_x; ++j, line <<= 1)
378+ if (line & 0x80) switch (orientation) {
379+ case 0: pixel(x + j, y - i) = c; break;
380+ case 1: pixel(x + i, y + j) = c; break;
381+ }
382+ }
383+ switch (orientation) {
384+ case 0: x += font_X; break;
385+ case 1: y += font_X; break;
386+ }
387+ }
388+}
389+
390+static int axis(double to, int max_steps, int * limit, char * * prefix)
391+{
392+ double scale = 1, step = 1;
393+ int i, prefix_num = 0;
394+ if (max_steps) {
395+ double try, log_10 = HUGE_VAL, min_step = (to *= 10) / max_steps;
396+ for (i = 5; i; i >>= 1) if ((try = ceil(log10(min_step * i))) <= log_10)
397+ step = pow(10., log_10 = try) / i, log_10 -= i > 1;
398+ prefix_num = floor(log_10 / 3);
399+ scale = pow(10., -3. * prefix_num);
400+ }
401+ *prefix = "pnum-kMGTPE" + prefix_num + (prefix_num? 4 : 11);
402+ *limit = to * scale + .5;
403+ return step * scale + .5;
404+}
405+
406+#define below 48
407+#define left 58
408+#define between 37
409+#define spectrum_width 14
410+#define right 35
411+
412+static int stop(sox_effect_t * effp)
413+{
414+ priv_t * p = (priv_t *) effp->priv;
415+ FILE * file = fopen(p->out_name, "w");
416+ uLong font_len = 96 * font_y;
417+ int rows = below + p->rows + 30 + 20 * !!p->title;
418+ int cols = left + p->cols + between + spectrum_width + right;
419+ sox_bool do_frame = !p->no_axes;
420+ png_byte * pixels = malloc(cols * rows * sizeof(*pixels));
421+ png_bytepp png_rows = malloc(rows * sizeof(*png_rows));
422+ png_structp png = png_create_write_struct(PNG_LIBPNG_VER_STRING, 0, 0,0);
423+ png_infop png_info = png_create_info_struct(png);
424+ png_byte palette[palette_size * 3];
425+ int i, j, step, limit, tick_len = 2 + do_frame;
426+ char text[200], * prefix;
427+
428+ if (!file) {
429+ sox_fail("failed to create `%s' :(", p->out_name);
430+ png_destroy_write_struct(&png, &png_info);
431+ free(png_rows);
432+ free(pixels);
433+ free(p->dBfs);
434+ return SOX_EOF;
435+ }
436+ sox_debug("signal-max=%g", p->max);
437+ font = malloc(font_len);
438+ assert(uncompress(font, &font_len, fixed, sizeof(fixed)-1) == Z_OK);
439+ make_palette(palette, p->style);
440+ memset(pixels, Background, cols * rows * sizeof(*pixels));
441+ png_init_io(png, file);
442+ png_set_PLTE(png, png_info, (png_colorp)palette, palette_size);
443+ png_set_IHDR(png, png_info, (size_t)cols, (size_t)rows, 8,
444+ PNG_COLOR_TYPE_PALETTE, PNG_INTERLACE_NONE,
445+ PNG_COMPRESSION_TYPE_DEFAULT, PNG_FILTER_TYPE_DEFAULT);
446+ for (j = 0; j < rows; ++j)
447+ png_rows[rows - 1 - j] = (png_bytep)(pixels + j * cols);
448+
449+ if (p->title && (i = (int)strlen(p->title) * font_X) < cols + 1) /* Title */
450+ print_at((cols - i) / 2, rows - font_y, Text, p->title);
451+ if ((int)strlen(p->comment) * font_X < cols + 1) /* Footer comment */
452+ print_at(1, font_y, Text, p->comment);
453+
454+ /* Spectrogram */
455+ for (j = 0; j < p->rows; ++j) {
456+ for (i = 0; i < p->cols; ++i)
457+ pixel(left + i, below + j) = colour(p->dBfs[i*p->rows + j], p->dB_range);
458+ if (do_frame) /* Frame verticals */
459+ pixel(left - 1, below + j) = pixel(left + p->cols,below + j) = Grid;
460+ }
461+ if (do_frame) for (i = -1; i <= p->cols; ++i) /* Frame horizontals */
462+ pixel(left + i, below - 1) = pixel(left + i, below + p->rows) = Grid;
463+
464+ /* X-axis */
465+ step = axis((double)p->cols * p->step_size * p->block_steps / effp->in_signal.rate, p->cols / (font_X * 9 / 2), &limit, &prefix);
466+ sprintf(text, "Time (%.1ss)", prefix); /* Axis label */
467+ print_at(left + (p->cols - font_X * (int)strlen(text)) / 2, 24, Text, text);
468+ for (i = 0; i <= limit; i += step) {
469+ int y, x = limit? (double)i / limit * p->cols + .5 : 0;
470+ for (y = 0; y < tick_len; ++y) /* Ticks */
471+ pixel(left - 1 + x, below - 1 - y) = pixel(left - 1 + x, below + p->rows + y) = Grid;
472+ if (step == 5 && (i%10))
473+ continue;
474+ sprintf(text, "%g", .1 * i); /* Tick labels */
475+ x = left + x - 3 * strlen(text);
476+ print_at(x, below - 6, Labels, text);
477+ print_at(x, below + p->rows + 14, Labels, text);
478+ }
479+
480+ /* Y-axis */
481+ step = axis(effp->in_signal.rate / 2, (p->rows - 1) / ((font_y * 3 + 1) >> 1), &limit, &prefix);
482+ sprintf(text, "Frequency (%.1sHz)", prefix); /* Axis label */
483+ print_up(10, below + (p->rows - font_X * (int)strlen(text)) / 2, Text, text);
484+ for (i = 0; i <= limit; i += step) {
485+ int x, y = limit? (double)i / limit * (p->rows - 1) + .5 : 0;
486+ for (x = 0; x < tick_len; ++x) /* Ticks */
487+ pixel(left - 1 - x, below + y) = pixel(left + p->cols + x, below + y) = Grid;
488+ if (step == 5 && (i%10))
489+ continue;
490+ sprintf(text, i?"%5g":" DC", .1 * i); /* Tick labels */
491+ print_at(left - 4 - font_X * 5, below + y + 5, Labels, text);
492+ sprintf(text, i?"%g":"DC", .1 * i);
493+ print_at(left + p->cols + 6, below + y + 5, Labels, text);
494+ }
495+
496+ /* Z-axis */
497+ print_at(cols - right - 2 - font_X, below - 13, Text, "dBFS");/* Axis label */
498+ for (j = 0; j < p->rows; ++j) { /* Spectrum */
499+ png_byte b = colour(p->dB_range * (j / (p->rows - 1.) - 1), p->dB_range);
500+ for (i = 0; i < spectrum_width; ++i)
501+ pixel(cols - right - 1 - i, below + j) = b;
502+ }
503+ for (i = 0; i <= p->dB_range; i += 10) { /* (Tick) labels */
504+ int y = (double)i / p->dB_range * (p->rows - 1) + .5;
505+ sprintf(text, "%+i", i - p->gain - p->dB_range);
506+ print_at(cols - right + 1, below + y + 5, Labels, text);
507+ }
508+
509+ free(font);
510+ png_set_rows(png, png_info, png_rows);
511+ png_write_png(png, png_info, PNG_TRANSFORM_IDENTITY, NULL);
512+ png_destroy_write_struct(&png, &png_info);
513+ free(png_rows);
514+ free(pixels);
515+ fclose(file);
516+ free(p->dBfs);
517+ return SOX_SUCCESS;
518+}
519+
520+sox_effect_handler_t const * sox_spectrogram_effect_fn(void)
521+{
522+ static sox_effect_handler_t handler = {
523+ "spectrogram", "[options]\n"
524+ " -x num X-axis pixels/second, default 100\n"
525+ " -y num Y-axis resolution (1 - 4), default 2\n"
526+ " -z num Z-axis (colour) range in dB, default 120\n"
527+ " -g num Apply num dB gain; to shift Z-axis\n"
528+ " -w name Window: Hann (default), Hamming, Bartlett, Rectangular, Kaiser\n"
529+ " -s num Display style (1 - 4), default 1\n"
530+ " -a Toggle the presence of the axes' lines\n"
531+ " -t text Title text\n"
532+ " -c text Comment text\n"
533+ " -o text Name of the spectrogram output file, default `spectrogram.png'\n",
534+ 0, getopts, start, flow, drain, stop, NULL, sizeof(priv_t)};
535+ return &handler;
536+}

Keyboard shortcuts

General

Show keyboard shortcuts
?
Focus the current search
/

Issues and pull requests

Open filters
f

Issues, pull requests and Inbox

Select next item
j↓
Select previous item
k↑
Open selected item
Enter
Dismiss filters or clear selection
Esc

Repository code

Go to file
t