/src/aac/libSACenc/src/sacenc_onsetdetect.cpp
Line | Count | Source (jump to first uncovered line) |
1 | | /* ----------------------------------------------------------------------------- |
2 | | Software License for The Fraunhofer FDK AAC Codec Library for Android |
3 | | |
4 | | © Copyright 1995 - 2018 Fraunhofer-Gesellschaft zur Förderung der angewandten |
5 | | Forschung e.V. All rights reserved. |
6 | | |
7 | | 1. INTRODUCTION |
8 | | The Fraunhofer FDK AAC Codec Library for Android ("FDK AAC Codec") is software |
9 | | that implements the MPEG Advanced Audio Coding ("AAC") encoding and decoding |
10 | | scheme for digital audio. This FDK AAC Codec software is intended to be used on |
11 | | a wide variety of Android devices. |
12 | | |
13 | | AAC's HE-AAC and HE-AAC v2 versions are regarded as today's most efficient |
14 | | general perceptual audio codecs. AAC-ELD is considered the best-performing |
15 | | full-bandwidth communications codec by independent studies and is widely |
16 | | deployed. AAC has been standardized by ISO and IEC as part of the MPEG |
17 | | specifications. |
18 | | |
19 | | Patent licenses for necessary patent claims for the FDK AAC Codec (including |
20 | | those of Fraunhofer) may be obtained through Via Licensing |
21 | | (www.vialicensing.com) or through the respective patent owners individually for |
22 | | the purpose of encoding or decoding bit streams in products that are compliant |
23 | | with the ISO/IEC MPEG audio standards. Please note that most manufacturers of |
24 | | Android devices already license these patent claims through Via Licensing or |
25 | | directly from the patent owners, and therefore FDK AAC Codec software may |
26 | | already be covered under those patent licenses when it is used for those |
27 | | licensed purposes only. |
28 | | |
29 | | Commercially-licensed AAC software libraries, including floating-point versions |
30 | | with enhanced sound quality, are also available from Fraunhofer. Users are |
31 | | encouraged to check the Fraunhofer website for additional applications |
32 | | information and documentation. |
33 | | |
34 | | 2. COPYRIGHT LICENSE |
35 | | |
36 | | Redistribution and use in source and binary forms, with or without modification, |
37 | | are permitted without payment of copyright license fees provided that you |
38 | | satisfy the following conditions: |
39 | | |
40 | | You must retain the complete text of this software license in redistributions of |
41 | | the FDK AAC Codec or your modifications thereto in source code form. |
42 | | |
43 | | You must retain the complete text of this software license in the documentation |
44 | | and/or other materials provided with redistributions of the FDK AAC Codec or |
45 | | your modifications thereto in binary form. You must make available free of |
46 | | charge copies of the complete source code of the FDK AAC Codec and your |
47 | | modifications thereto to recipients of copies in binary form. |
48 | | |
49 | | The name of Fraunhofer may not be used to endorse or promote products derived |
50 | | from this library without prior written permission. |
51 | | |
52 | | You may not charge copyright license fees for anyone to use, copy or distribute |
53 | | the FDK AAC Codec software or your modifications thereto. |
54 | | |
55 | | Your modified versions of the FDK AAC Codec must carry prominent notices stating |
56 | | that you changed the software and the date of any change. For modified versions |
57 | | of the FDK AAC Codec, the term "Fraunhofer FDK AAC Codec Library for Android" |
58 | | must be replaced by the term "Third-Party Modified Version of the Fraunhofer FDK |
59 | | AAC Codec Library for Android." |
60 | | |
61 | | 3. NO PATENT LICENSE |
62 | | |
63 | | NO EXPRESS OR IMPLIED LICENSES TO ANY PATENT CLAIMS, including without |
64 | | limitation the patents of Fraunhofer, ARE GRANTED BY THIS SOFTWARE LICENSE. |
65 | | Fraunhofer provides no warranty of patent non-infringement with respect to this |
66 | | software. |
67 | | |
68 | | You may use this FDK AAC Codec software or modifications thereto only for |
69 | | purposes that are authorized by appropriate patent licenses. |
70 | | |
71 | | 4. DISCLAIMER |
72 | | |
73 | | This FDK AAC Codec software is provided by Fraunhofer on behalf of the copyright |
74 | | holders and contributors "AS IS" and WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, |
75 | | including but not limited to the implied warranties of merchantability and |
76 | | fitness for a particular purpose. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR |
77 | | CONTRIBUTORS BE LIABLE for any direct, indirect, incidental, special, exemplary, |
78 | | or consequential damages, including but not limited to procurement of substitute |
79 | | goods or services; loss of use, data, or profits, or business interruption, |
80 | | however caused and on any theory of liability, whether in contract, strict |
81 | | liability, or tort (including negligence), arising in any way out of the use of |
82 | | this software, even if advised of the possibility of such damage. |
83 | | |
84 | | 5. CONTACT INFORMATION |
85 | | |
86 | | Fraunhofer Institute for Integrated Circuits IIS |
87 | | Attention: Audio and Multimedia Departments - FDK AAC LL |
88 | | Am Wolfsmantel 33 |
89 | | 91058 Erlangen, Germany |
90 | | |
91 | | www.iis.fraunhofer.de/amm |
92 | | amm-info@iis.fraunhofer.de |
93 | | ----------------------------------------------------------------------------- */ |
94 | | |
95 | | /*********************** MPEG surround encoder library ************************* |
96 | | |
97 | | Author(s): Max Neuendorf |
98 | | |
99 | | Description: Encoder Library Interface |
100 | | Detect Onset in current frame |
101 | | |
102 | | *******************************************************************************/ |
103 | | |
104 | | /**************************************************************************/ /** |
105 | | \file |
106 | | Description of file contents |
107 | | ******************************************************************************/ |
108 | | |
109 | | /* Includes ******************************************************************/ |
110 | | #include "sacenc_onsetdetect.h" |
111 | | #include "genericStds.h" |
112 | | #include "sacenc_vectorfunctions.h" |
113 | | |
114 | | /* Defines *******************************************************************/ |
115 | | #define SPACE_ONSET_THRESHOLD (3.0) |
116 | 0 | #define SPACE_ONSET_THRESHOLD_SF (3) |
117 | | #define SPACE_ONSET_THRESHOLD_SQUARE \ |
118 | 0 | (FL2FXCONST_DBL((1.0 / (SPACE_ONSET_THRESHOLD * SPACE_ONSET_THRESHOLD)) * \ |
119 | 0 | (float)(1 << SPACE_ONSET_THRESHOLD_SF))) |
120 | | |
121 | | /* Data Types ****************************************************************/ |
122 | | struct ONSET_DETECT { |
123 | | INT maxTimeSlots; |
124 | | INT minTransientDistance; |
125 | | INT avgEnergyDistance; |
126 | | INT lowerBoundOnsetDetection; |
127 | | INT upperBoundOnsetDetection; |
128 | | FIXP_DBL *pEnergyHist__FDK; |
129 | | SCHAR *pEnergyHistScale; |
130 | | SCHAR avgEnergyDistanceScale; |
131 | | }; |
132 | | |
133 | | /* Constants *****************************************************************/ |
134 | | |
135 | | /* Function / Class Declarations *********************************************/ |
136 | | |
137 | | /* Function / Class Definition ***********************************************/ |
138 | | FDK_SACENC_ERROR fdk_sacenc_onsetDetect_Open(HANDLE_ONSET_DETECT *phOnset, |
139 | 0 | const UINT maxTimeSlots) { |
140 | 0 | FDK_SACENC_ERROR error = SACENC_OK; |
141 | 0 | HANDLE_ONSET_DETECT hOnset = NULL; |
142 | |
|
143 | 0 | if (NULL == phOnset) { |
144 | 0 | error = SACENC_INVALID_HANDLE; |
145 | 0 | } else { |
146 | | /* Memory Allocation */ |
147 | 0 | FDK_ALLOCATE_MEMORY_1D(hOnset, 1, struct ONSET_DETECT); |
148 | 0 | FDK_ALLOCATE_MEMORY_1D(hOnset->pEnergyHist__FDK, 16 + maxTimeSlots, |
149 | 0 | FIXP_DBL); |
150 | 0 | FDK_ALLOCATE_MEMORY_1D(hOnset->pEnergyHistScale, 16 + maxTimeSlots, SCHAR); |
151 | |
|
152 | 0 | hOnset->maxTimeSlots = maxTimeSlots; |
153 | 0 | hOnset->minTransientDistance = |
154 | 0 | 8; /* minimum distance between detected transients */ |
155 | 0 | hOnset->avgEnergyDistance = 16; /* average energy distance */ |
156 | |
|
157 | 0 | hOnset->avgEnergyDistanceScale = 4; |
158 | 0 | *phOnset = hOnset; |
159 | 0 | } |
160 | 0 | return error; |
161 | | |
162 | 0 | bail: |
163 | 0 | fdk_sacenc_onsetDetect_Close(&hOnset); |
164 | 0 | return ((SACENC_OK == error) ? SACENC_MEMORY_ERROR : error); |
165 | 0 | } |
166 | | |
167 | | FDK_SACENC_ERROR fdk_sacenc_onsetDetect_Init( |
168 | | HANDLE_ONSET_DETECT hOnset, |
169 | 0 | const ONSET_DETECT_CONFIG *const pOnsetDetectConfig, const UINT initFlags) { |
170 | 0 | FDK_SACENC_ERROR error = SACENC_OK; |
171 | |
|
172 | 0 | if ((NULL == hOnset) || (pOnsetDetectConfig == NULL)) { |
173 | 0 | error = SACENC_INVALID_HANDLE; |
174 | 0 | } else { |
175 | 0 | if ((pOnsetDetectConfig->maxTimeSlots > hOnset->maxTimeSlots) || |
176 | 0 | (pOnsetDetectConfig->upperBoundOnsetDetection < |
177 | 0 | hOnset->lowerBoundOnsetDetection)) { |
178 | 0 | error = SACENC_INVALID_CONFIG; |
179 | 0 | goto bail; |
180 | 0 | } |
181 | | |
182 | 0 | hOnset->maxTimeSlots = pOnsetDetectConfig->maxTimeSlots; |
183 | 0 | hOnset->lowerBoundOnsetDetection = |
184 | 0 | pOnsetDetectConfig->lowerBoundOnsetDetection; |
185 | 0 | hOnset->upperBoundOnsetDetection = |
186 | 0 | pOnsetDetectConfig->upperBoundOnsetDetection; |
187 | |
|
188 | 0 | hOnset->minTransientDistance = |
189 | 0 | 8; /* minimum distance between detected transients */ |
190 | 0 | hOnset->avgEnergyDistance = 16; /* average energy distance */ |
191 | |
|
192 | 0 | hOnset->avgEnergyDistanceScale = 4; |
193 | | |
194 | | /* Init / Reset */ |
195 | 0 | if (initFlags) { |
196 | 0 | int i; |
197 | 0 | for (i = 0; i < hOnset->avgEnergyDistance + hOnset->maxTimeSlots; i++) |
198 | 0 | hOnset->pEnergyHistScale[i] = -(DFRACT_BITS - 3); |
199 | |
|
200 | 0 | FDKmemset_flex( |
201 | 0 | hOnset->pEnergyHist__FDK, |
202 | 0 | FL2FXCONST_DBL(SACENC_FLOAT_EPSILON * (1 << (DFRACT_BITS - 3))), |
203 | 0 | hOnset->avgEnergyDistance + hOnset->maxTimeSlots); |
204 | 0 | } |
205 | 0 | } |
206 | | |
207 | 0 | bail: |
208 | 0 | return error; |
209 | 0 | } |
210 | | |
211 | | /**************************************************************************/ |
212 | | |
213 | 0 | FDK_SACENC_ERROR fdk_sacenc_onsetDetect_Close(HANDLE_ONSET_DETECT *phOnset) { |
214 | 0 | FDK_SACENC_ERROR error = SACENC_OK; |
215 | |
|
216 | 0 | if ((NULL != phOnset) && (NULL != *phOnset)) { |
217 | 0 | if (NULL != (*phOnset)->pEnergyHist__FDK) { |
218 | 0 | FDKfree((*phOnset)->pEnergyHist__FDK); |
219 | 0 | } |
220 | 0 | (*phOnset)->pEnergyHist__FDK = NULL; |
221 | |
|
222 | 0 | if (NULL != (*phOnset)->pEnergyHistScale) { |
223 | 0 | FDKfree((*phOnset)->pEnergyHistScale); |
224 | 0 | } |
225 | 0 | (*phOnset)->pEnergyHistScale = NULL; |
226 | 0 | FDKfree(*phOnset); |
227 | 0 | *phOnset = NULL; |
228 | 0 | } |
229 | 0 | return error; |
230 | 0 | } |
231 | | |
232 | | /**************************************************************************/ |
233 | | |
234 | | FDK_SACENC_ERROR fdk_sacenc_onsetDetect_Update(HANDLE_ONSET_DETECT hOnset, |
235 | 0 | const INT timeSlots) { |
236 | 0 | FDK_SACENC_ERROR error = SACENC_OK; |
237 | |
|
238 | 0 | if (NULL == hOnset) { |
239 | 0 | error = SACENC_INVALID_HANDLE; |
240 | 0 | } else { |
241 | 0 | if (timeSlots > hOnset->maxTimeSlots) { |
242 | 0 | error = SACENC_INVALID_CONFIG; |
243 | 0 | } else { |
244 | 0 | int i; |
245 | | /* Shift old data */ |
246 | 0 | for (i = 0; i < hOnset->avgEnergyDistance; i++) { |
247 | 0 | hOnset->pEnergyHist__FDK[i] = hOnset->pEnergyHist__FDK[i + timeSlots]; |
248 | 0 | hOnset->pEnergyHistScale[i] = hOnset->pEnergyHistScale[i + timeSlots]; |
249 | 0 | } |
250 | | |
251 | | /* Clear for new data */ |
252 | 0 | FDKmemset_flex(&hOnset->pEnergyHist__FDK[hOnset->avgEnergyDistance], |
253 | 0 | FL2FXCONST_DBL(SACENC_FLOAT_EPSILON), timeSlots); |
254 | 0 | } |
255 | 0 | } |
256 | 0 | return error; |
257 | 0 | } |
258 | | |
259 | | /**************************************************************************/ |
260 | | |
261 | | FDK_SACENC_ERROR fdk_sacenc_onsetDetect_Apply( |
262 | | HANDLE_ONSET_DETECT hOnset, const INT nTimeSlots, const INT nHybridBands, |
263 | | FIXP_DPK *const *const ppHybridData__FDK, const INT hybridDataScale, |
264 | 0 | const INT prevPos, INT pTransientPos[MAX_NUM_TRANS]) { |
265 | 0 | FDK_SACENC_ERROR error = SACENC_OK; |
266 | |
|
267 | 0 | C_ALLOC_SCRATCH_START(envs, FIXP_DBL, (16 + MAX_TIME_SLOTS)) |
268 | 0 | FDKmemclear(envs, (16 + MAX_TIME_SLOTS) * sizeof(FIXP_DBL)); |
269 | |
|
270 | 0 | if ((hOnset == NULL) || (pTransientPos == NULL) || |
271 | 0 | (ppHybridData__FDK == NULL)) { |
272 | 0 | error = SACENC_INVALID_HANDLE; |
273 | 0 | } else { |
274 | 0 | int i, ts, trCnt, currPos; |
275 | |
|
276 | 0 | if ((nTimeSlots < 0) || (nTimeSlots > hOnset->maxTimeSlots) || |
277 | 0 | (hOnset->lowerBoundOnsetDetection < -1) || |
278 | 0 | (hOnset->upperBoundOnsetDetection > nHybridBands)) { |
279 | 0 | error = SACENC_INVALID_CONFIG; |
280 | 0 | goto bail; |
281 | 0 | } |
282 | | |
283 | 0 | const int lowerBoundOnsetDetection = hOnset->lowerBoundOnsetDetection; |
284 | 0 | const int upperBoundOnsetDetection = hOnset->upperBoundOnsetDetection; |
285 | 0 | const int M = hOnset->avgEnergyDistance; |
286 | |
|
287 | 0 | { |
288 | 0 | SCHAR *envScale = hOnset->pEnergyHistScale; |
289 | 0 | FIXP_DBL *env = hOnset->pEnergyHist__FDK; |
290 | 0 | const FIXP_DBL threshold_square = SPACE_ONSET_THRESHOLD_SQUARE; |
291 | |
|
292 | 0 | trCnt = 0; |
293 | | |
294 | | /* reset transient array */ |
295 | 0 | FDKmemset_flex(pTransientPos, -1, MAX_NUM_TRANS); |
296 | | |
297 | | /* minimum transient distance of minTransDist QMF samples */ |
298 | 0 | if (prevPos > 0) { |
299 | 0 | currPos = FDKmax(nTimeSlots, |
300 | 0 | prevPos - nTimeSlots + hOnset->minTransientDistance); |
301 | 0 | } else { |
302 | 0 | currPos = nTimeSlots; |
303 | 0 | } |
304 | | |
305 | | /* get energy and scalefactor for each time slot */ |
306 | 0 | int outScale; |
307 | 0 | int inScale = 3; /* scale factor determined empirically */ |
308 | 0 | for (ts = 0; ts < nTimeSlots; ts++) { |
309 | 0 | env[M + ts] = sumUpCplxPow2( |
310 | 0 | &ppHybridData__FDK[ts][lowerBoundOnsetDetection + 1], |
311 | 0 | SUM_UP_DYNAMIC_SCALE, inScale, &outScale, |
312 | 0 | upperBoundOnsetDetection - lowerBoundOnsetDetection - 1); |
313 | 0 | envScale[M + ts] = outScale + (hybridDataScale << 1); |
314 | 0 | } |
315 | | |
316 | | /* calculate common scale for all time slots */ |
317 | 0 | SCHAR maxScale = -(DFRACT_BITS - 1); |
318 | 0 | for (i = 0; i < (nTimeSlots + M); i++) { |
319 | 0 | maxScale = fixMax(maxScale, envScale[i]); |
320 | 0 | } |
321 | | |
322 | | /* apply common scale and store energy in temporary buffer */ |
323 | 0 | for (i = 0; i < (nTimeSlots + M); i++) { |
324 | 0 | envs[i] = env[i] >> fixMin((maxScale - envScale[i]), (DFRACT_BITS - 1)); |
325 | 0 | } |
326 | |
|
327 | 0 | FIXP_DBL maxVal = FL2FXCONST_DBL(0.0f); |
328 | 0 | for (i = 0; i < (nTimeSlots + M); i++) { |
329 | 0 | maxVal |= fAbs(envs[i]); |
330 | 0 | } |
331 | |
|
332 | 0 | int s = fixMax(0, CntLeadingZeros(maxVal) - 1); |
333 | |
|
334 | 0 | for (i = 0; i < (nTimeSlots + M); i++) { |
335 | 0 | envs[i] = envs[i] << s; |
336 | 0 | } |
337 | |
|
338 | 0 | int currPosPrev = currPos; |
339 | 0 | FIXP_DBL p1, p2; |
340 | 0 | p2 = FL2FXCONST_DBL(0.0f); |
341 | 0 | for (; (currPos < (nTimeSlots << 1)) && (trCnt < MAX_NUM_TRANS); |
342 | 0 | currPos++) { |
343 | 0 | p1 = fMultDiv2(envs[currPos - nTimeSlots + M], threshold_square) >> |
344 | 0 | (SPACE_ONSET_THRESHOLD_SF - 1); |
345 | | |
346 | | /* Calculate average of past M energy values */ |
347 | 0 | if (currPosPrev == (currPos - 1)) { |
348 | | /* remove last and add new element */ |
349 | 0 | p2 -= (envs[currPosPrev - nTimeSlots] >> |
350 | 0 | (int)hOnset->avgEnergyDistanceScale); |
351 | 0 | p2 += (envs[currPos - nTimeSlots + M - 1] >> |
352 | 0 | (int)hOnset->avgEnergyDistanceScale); |
353 | 0 | } else { |
354 | | /* calculate complete vector */ |
355 | 0 | p2 = FL2FXCONST_DBL(0.0f); |
356 | 0 | for (ts = 0; ts < M; ts++) { |
357 | 0 | p2 += (envs[currPos - nTimeSlots + ts] >> |
358 | 0 | (int)hOnset->avgEnergyDistanceScale); |
359 | 0 | } |
360 | 0 | } |
361 | 0 | currPosPrev = currPos; |
362 | |
|
363 | 0 | { |
364 | | /* save position if transient found */ |
365 | 0 | if (p1 > p2) { |
366 | 0 | pTransientPos[trCnt++] = currPos; |
367 | 0 | currPos += hOnset->minTransientDistance; |
368 | 0 | } |
369 | 0 | } |
370 | 0 | } /* for currPos */ |
371 | 0 | } |
372 | |
|
373 | 0 | } /* valid handle*/ |
374 | 0 | bail: |
375 | |
|
376 | 0 | C_ALLOC_SCRATCH_END(envs, FIXP_DBL, (16 + MAX_TIME_SLOTS)) |
377 | |
|
378 | 0 | return error; |
379 | 0 | } |
380 | | |
381 | | /**************************************************************************/ |