mbed official / mbed-dsp

CMSIS DSP library

Dependents:   performance_timer Surfboard_ gps2rtty Capstone ... more

Legacy Warning

This is an mbed 2 library. To learn more about mbed OS 5, visit the docs.

cmsis_dsp/FilteringFunctions/arm_lms_norm_f32.c@5:3762170b6d4d, 2015-11-20 (annotated)

Committer:
mbed_official
Date:
Fri Nov 20 08:45:18 2015 +0000
Revision:
5:3762170b6d4d
Parent:
3:7a284390b0ce 
Synchronized with git revision 2eb940b9a73af188d3004a2575fdfbb05febe62b



Full URL: https://github.com/mbedmicro/mbed/commit/2eb940b9a73af188d3004a2575fdfbb05febe62b/



Added option to build rpc library. closes #1426

Who changed what in which revision?

UserRevisionLine numberNew contents of line
emilmont 1:fdd22bb7aa52 1 /* ----------------------------------------------------------------------
mbed_official 5:3762170b6d4d 2 * Copyright (C) 2010-2014 ARM Limited. All rights reserved.
emilmont 1:fdd22bb7aa52 3 *
mbed_official 5:3762170b6d4d 4 * $Date: 19. March 2015
mbed_official 5:3762170b6d4d 5 * $Revision: V.1.4.5
emilmont 1:fdd22bb7aa52 6 *
emilmont 2:da51fb522205 7 * Project: CMSIS DSP Library
emilmont 2:da51fb522205 8 * Title: arm_lms_norm_f32.c
emilmont 1:fdd22bb7aa52 9 *
emilmont 2:da51fb522205 10 * Description: Processing function for the floating-point Normalised LMS.
emilmont 1:fdd22bb7aa52 11 *
emilmont 1:fdd22bb7aa52 12 * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0
emilmont 1:fdd22bb7aa52 13 *
mbed_official 3:7a284390b0ce 14 * Redistribution and use in source and binary forms, with or without
mbed_official 3:7a284390b0ce 15 * modification, are permitted provided that the following conditions
mbed_official 3:7a284390b0ce 16 * are met:
mbed_official 3:7a284390b0ce 17 * - Redistributions of source code must retain the above copyright
mbed_official 3:7a284390b0ce 18 * notice, this list of conditions and the following disclaimer.
mbed_official 3:7a284390b0ce 19 * - Redistributions in binary form must reproduce the above copyright
mbed_official 3:7a284390b0ce 20 * notice, this list of conditions and the following disclaimer in
mbed_official 3:7a284390b0ce 21 * the documentation and/or other materials provided with the
mbed_official 3:7a284390b0ce 22 * distribution.
mbed_official 3:7a284390b0ce 23 * - Neither the name of ARM LIMITED nor the names of its contributors
mbed_official 3:7a284390b0ce 24 * may be used to endorse or promote products derived from this
mbed_official 3:7a284390b0ce 25 * software without specific prior written permission.
mbed_official 3:7a284390b0ce 26 *
mbed_official 3:7a284390b0ce 27 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
mbed_official 3:7a284390b0ce 28 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
mbed_official 3:7a284390b0ce 29 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
mbed_official 3:7a284390b0ce 30 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
mbed_official 3:7a284390b0ce 31 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
mbed_official 3:7a284390b0ce 32 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
mbed_official 3:7a284390b0ce 33 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
mbed_official 3:7a284390b0ce 34 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
mbed_official 3:7a284390b0ce 35 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
mbed_official 3:7a284390b0ce 36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
mbed_official 3:7a284390b0ce 37 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
mbed_official 3:7a284390b0ce 38 * POSSIBILITY OF SUCH DAMAGE.
emilmont 1:fdd22bb7aa52 39 * -------------------------------------------------------------------- */
emilmont 1:fdd22bb7aa52 40
emilmont 1:fdd22bb7aa52 41 #include "arm_math.h"
emilmont 1:fdd22bb7aa52 42
emilmont 1:fdd22bb7aa52 43 /**
emilmont 1:fdd22bb7aa52 44 * @ingroup groupFilters
emilmont 1:fdd22bb7aa52 45 */
emilmont 1:fdd22bb7aa52 46
emilmont 1:fdd22bb7aa52 47 /**
emilmont 1:fdd22bb7aa52 48 * @defgroup LMS_NORM Normalized LMS Filters
emilmont 1:fdd22bb7aa52 49 *
emilmont 1:fdd22bb7aa52 50 * This set of functions implements a commonly used adaptive filter.
emilmont 1:fdd22bb7aa52 51 * It is related to the Least Mean Square (LMS) adaptive filter and includes an additional normalization
emilmont 1:fdd22bb7aa52 52 * factor which increases the adaptation rate of the filter.
emilmont 1:fdd22bb7aa52 53 * The CMSIS DSP Library contains normalized LMS filter functions that operate on Q15, Q31, and floating-point data types.
emilmont 1:fdd22bb7aa52 54 *
emilmont 1:fdd22bb7aa52 55 * A normalized least mean square (NLMS) filter consists of two components as shown below.
emilmont 1:fdd22bb7aa52 56 * The first component is a standard transversal or FIR filter.
emilmont 1:fdd22bb7aa52 57 * The second component is a coefficient update mechanism.
emilmont 1:fdd22bb7aa52 58 * The NLMS filter has two input signals.
emilmont 1:fdd22bb7aa52 59 * The "input" feeds the FIR filter while the "reference input" corresponds to the desired output of the FIR filter.
emilmont 1:fdd22bb7aa52 60 * That is, the FIR filter coefficients are updated so that the output of the FIR filter matches the reference input.
emilmont 1:fdd22bb7aa52 61 * The filter coefficient update mechanism is based on the difference between the FIR filter output and the reference input.
emilmont 1:fdd22bb7aa52 62 * This "error signal" tends towards zero as the filter adapts.
emilmont 1:fdd22bb7aa52 63 * The NLMS processing functions accept the input and reference input signals and generate the filter output and error signal.
emilmont 1:fdd22bb7aa52 64 * \image html LMS.gif "Internal structure of the NLMS adaptive filter"
emilmont 1:fdd22bb7aa52 65 *
emilmont 1:fdd22bb7aa52 66 * The functions operate on blocks of data and each call to the function processes
emilmont 1:fdd22bb7aa52 67 * <code>blockSize</code> samples through the filter.
emilmont 1:fdd22bb7aa52 68 * <code>pSrc</code> points to input signal, <code>pRef</code> points to reference signal,
emilmont 1:fdd22bb7aa52 69 * <code>pOut</code> points to output signal and <code>pErr</code> points to error signal.
emilmont 1:fdd22bb7aa52 70 * All arrays contain <code>blockSize</code> values.
emilmont 1:fdd22bb7aa52 71 *
emilmont 1:fdd22bb7aa52 72 * The functions operate on a block-by-block basis.
emilmont 1:fdd22bb7aa52 73 * Internally, the filter coefficients <code>b[n]</code> are updated on a sample-by-sample basis.
emilmont 1:fdd22bb7aa52 74 * The convergence of the LMS filter is slower compared to the normalized LMS algorithm.
emilmont 1:fdd22bb7aa52 75 *
emilmont 1:fdd22bb7aa52 76 * \par Algorithm:
emilmont 1:fdd22bb7aa52 77 * The output signal <code>y[n]</code> is computed by a standard FIR filter:
emilmont 1:fdd22bb7aa52 78 * <pre>
emilmont 1:fdd22bb7aa52 79 * y[n] = b[0] * x[n] + b[1] * x[n-1] + b[2] * x[n-2] + ...+ b[numTaps-1] * x[n-numTaps+1]
emilmont 1:fdd22bb7aa52 80 * </pre>
emilmont 1:fdd22bb7aa52 81 *
emilmont 1:fdd22bb7aa52 82 * \par
emilmont 1:fdd22bb7aa52 83 * The error signal equals the difference between the reference signal <code>d[n]</code> and the filter output:
emilmont 1:fdd22bb7aa52 84 * <pre>
emilmont 1:fdd22bb7aa52 85 * e[n] = d[n] - y[n].
emilmont 1:fdd22bb7aa52 86 * </pre>
emilmont 1:fdd22bb7aa52 87 *
emilmont 1:fdd22bb7aa52 88 * \par
emilmont 1:fdd22bb7aa52 89 * After each sample of the error signal is computed the instanteous energy of the filter state variables is calculated:
emilmont 1:fdd22bb7aa52 90 * <pre>
emilmont 1:fdd22bb7aa52 91 * E = x[n]^2 + x[n-1]^2 + ... + x[n-numTaps+1]^2.
emilmont 1:fdd22bb7aa52 92 * </pre>
emilmont 1:fdd22bb7aa52 93 * The filter coefficients <code>b[k]</code> are then updated on a sample-by-sample basis:
emilmont 1:fdd22bb7aa52 94 * <pre>
emilmont 1:fdd22bb7aa52 95 * b[k] = b[k] + e[n] * (mu/E) * x[n-k], for k=0, 1, ..., numTaps-1
emilmont 1:fdd22bb7aa52 96 * </pre>
emilmont 1:fdd22bb7aa52 97 * where <code>mu</code> is the step size and controls the rate of coefficient convergence.
emilmont 1:fdd22bb7aa52 98 *\par
emilmont 1:fdd22bb7aa52 99 * In the APIs, <code>pCoeffs</code> points to a coefficient array of size <code>numTaps</code>.
emilmont 1:fdd22bb7aa52 100 * Coefficients are stored in time reversed order.
emilmont 1:fdd22bb7aa52 101 * \par
emilmont 1:fdd22bb7aa52 102 * <pre>
emilmont 1:fdd22bb7aa52 103 * {b[numTaps-1], b[numTaps-2], b[N-2], ..., b[1], b[0]}
emilmont 1:fdd22bb7aa52 104 * </pre>
emilmont 1:fdd22bb7aa52 105 * \par
emilmont 1:fdd22bb7aa52 106 * <code>pState</code> points to a state array of size <code>numTaps + blockSize - 1</code>.
emilmont 1:fdd22bb7aa52 107 * Samples in the state buffer are stored in the order:
emilmont 1:fdd22bb7aa52 108 * \par
emilmont 1:fdd22bb7aa52 109 * <pre>
emilmont 1:fdd22bb7aa52 110 * {x[n-numTaps+1], x[n-numTaps], x[n-numTaps-1], x[n-numTaps-2]....x[0], x[1], ..., x[blockSize-1]}
emilmont 1:fdd22bb7aa52 111 * </pre>
emilmont 1:fdd22bb7aa52 112 * \par
emilmont 1:fdd22bb7aa52 113 * Note that the length of the state buffer exceeds the length of the coefficient array by <code>blockSize-1</code> samples.
emilmont 1:fdd22bb7aa52 114 * The increased state buffer length allows circular addressing, which is traditionally used in FIR filters,
emilmont 1:fdd22bb7aa52 115 * to be avoided and yields a significant speed improvement.
emilmont 1:fdd22bb7aa52 116 * The state variables are updated after each block of data is processed.
emilmont 1:fdd22bb7aa52 117 * \par Instance Structure
emilmont 1:fdd22bb7aa52 118 * The coefficients and state variables for a filter are stored together in an instance data structure.
emilmont 1:fdd22bb7aa52 119 * A separate instance structure must be defined for each filter and
emilmont 1:fdd22bb7aa52 120 * coefficient and state arrays cannot be shared among instances.
emilmont 1:fdd22bb7aa52 121 * There are separate instance structure declarations for each of the 3 supported data types.
emilmont 1:fdd22bb7aa52 122 *
emilmont 1:fdd22bb7aa52 123 * \par Initialization Functions
emilmont 1:fdd22bb7aa52 124 * There is also an associated initialization function for each data type.
emilmont 1:fdd22bb7aa52 125 * The initialization function performs the following operations:
emilmont 1:fdd22bb7aa52 126 * - Sets the values of the internal structure fields.
emilmont 1:fdd22bb7aa52 127 * - Zeros out the values in the state buffer.
mbed_official 3:7a284390b0ce 128 * To do this manually without calling the init function, assign the follow subfields of the instance structure:
mbed_official 3:7a284390b0ce 129 * numTaps, pCoeffs, mu, energy, x0, pState. Also set all of the values in pState to zero.
mbed_official 3:7a284390b0ce 130 * For Q7, Q15, and Q31 the following fields must also be initialized;
mbed_official 3:7a284390b0ce 131 * recipTable, postShift
mbed_official 3:7a284390b0ce 132 *
emilmont 1:fdd22bb7aa52 133 * \par
emilmont 1:fdd22bb7aa52 134 * Instance structure cannot be placed into a const data section and it is recommended to use the initialization function.
emilmont 1:fdd22bb7aa52 135 * \par Fixed-Point Behavior:
emilmont 1:fdd22bb7aa52 136 * Care must be taken when using the Q15 and Q31 versions of the normalised LMS filter.
emilmont 1:fdd22bb7aa52 137 * The following issues must be considered:
emilmont 1:fdd22bb7aa52 138 * - Scaling of coefficients
emilmont 1:fdd22bb7aa52 139 * - Overflow and saturation
emilmont 1:fdd22bb7aa52 140 *
emilmont 1:fdd22bb7aa52 141 * \par Scaling of Coefficients:
emilmont 1:fdd22bb7aa52 142 * Filter coefficients are represented as fractional values and
emilmont 1:fdd22bb7aa52 143 * coefficients are restricted to lie in the range <code>[-1 +1)</code>.
emilmont 1:fdd22bb7aa52 144 * The fixed-point functions have an additional scaling parameter <code>postShift</code>.
emilmont 1:fdd22bb7aa52 145 * At the output of the filter's accumulator is a shift register which shifts the result by <code>postShift</code> bits.
emilmont 1:fdd22bb7aa52 146 * This essentially scales the filter coefficients by <code>2^postShift</code> and
emilmont 1:fdd22bb7aa52 147 * allows the filter coefficients to exceed the range <code>[+1 -1)</code>.
emilmont 1:fdd22bb7aa52 148 * The value of <code>postShift</code> is set by the user based on the expected gain through the system being modeled.
emilmont 1:fdd22bb7aa52 149 *
emilmont 1:fdd22bb7aa52 150 * \par Overflow and Saturation:
emilmont 1:fdd22bb7aa52 151 * Overflow and saturation behavior of the fixed-point Q15 and Q31 versions are
emilmont 1:fdd22bb7aa52 152 * described separately as part of the function specific documentation below.
emilmont 1:fdd22bb7aa52 153 */
emilmont 1:fdd22bb7aa52 154
emilmont 1:fdd22bb7aa52 155
emilmont 1:fdd22bb7aa52 156 /**
emilmont 1:fdd22bb7aa52 157 * @addtogroup LMS_NORM
emilmont 1:fdd22bb7aa52 158 * @{
emilmont 1:fdd22bb7aa52 159 */
emilmont 1:fdd22bb7aa52 160
emilmont 1:fdd22bb7aa52 161
emilmont 1:fdd22bb7aa52 162 /**
emilmont 1:fdd22bb7aa52 163 * @brief Processing function for floating-point normalized LMS filter.
emilmont 1:fdd22bb7aa52 164 * @param[in] *S points to an instance of the floating-point normalized LMS filter structure.
emilmont 1:fdd22bb7aa52 165 * @param[in] *pSrc points to the block of input data.
emilmont 1:fdd22bb7aa52 166 * @param[in] *pRef points to the block of reference data.
emilmont 1:fdd22bb7aa52 167 * @param[out] *pOut points to the block of output data.
emilmont 1:fdd22bb7aa52 168 * @param[out] *pErr points to the block of error data.
emilmont 1:fdd22bb7aa52 169 * @param[in] blockSize number of samples to process.
emilmont 1:fdd22bb7aa52 170 * @return none.
emilmont 1:fdd22bb7aa52 171 */
emilmont 1:fdd22bb7aa52 172
emilmont 1:fdd22bb7aa52 173 void arm_lms_norm_f32(
emilmont 1:fdd22bb7aa52 174 arm_lms_norm_instance_f32 * S,
emilmont 1:fdd22bb7aa52 175 float32_t * pSrc,
emilmont 1:fdd22bb7aa52 176 float32_t * pRef,
emilmont 1:fdd22bb7aa52 177 float32_t * pOut,
emilmont 1:fdd22bb7aa52 178 float32_t * pErr,
emilmont 1:fdd22bb7aa52 179 uint32_t blockSize)
emilmont 1:fdd22bb7aa52 180 {
emilmont 1:fdd22bb7aa52 181 float32_t *pState = S->pState; /* State pointer */
emilmont 1:fdd22bb7aa52 182 float32_t *pCoeffs = S->pCoeffs; /* Coefficient pointer */
emilmont 1:fdd22bb7aa52 183 float32_t *pStateCurnt; /* Points to the current sample of the state */
emilmont 1:fdd22bb7aa52 184 float32_t *px, *pb; /* Temporary pointers for state and coefficient buffers */
emilmont 1:fdd22bb7aa52 185 float32_t mu = S->mu; /* Adaptive factor */
emilmont 1:fdd22bb7aa52 186 uint32_t numTaps = S->numTaps; /* Number of filter coefficients in the filter */
emilmont 1:fdd22bb7aa52 187 uint32_t tapCnt, blkCnt; /* Loop counters */
emilmont 1:fdd22bb7aa52 188 float32_t energy; /* Energy of the input */
emilmont 1:fdd22bb7aa52 189 float32_t sum, e, d; /* accumulator, error, reference data sample */
emilmont 1:fdd22bb7aa52 190 float32_t w, x0, in; /* weight factor, temporary variable to hold input sample and state */
emilmont 1:fdd22bb7aa52 191
emilmont 1:fdd22bb7aa52 192 /* Initializations of error, difference, Coefficient update */
emilmont 1:fdd22bb7aa52 193 e = 0.0f;
emilmont 1:fdd22bb7aa52 194 d = 0.0f;
emilmont 1:fdd22bb7aa52 195 w = 0.0f;
emilmont 1:fdd22bb7aa52 196
emilmont 1:fdd22bb7aa52 197 energy = S->energy;
emilmont 1:fdd22bb7aa52 198 x0 = S->x0;
emilmont 1:fdd22bb7aa52 199
emilmont 1:fdd22bb7aa52 200 /* S->pState points to buffer which contains previous frame (numTaps - 1) samples */
emilmont 1:fdd22bb7aa52 201 /* pStateCurnt points to the location where the new input data should be written */
emilmont 1:fdd22bb7aa52 202 pStateCurnt = &(S->pState[(numTaps - 1u)]);
emilmont 1:fdd22bb7aa52 203
emilmont 1:fdd22bb7aa52 204 /* Loop over blockSize number of values */
emilmont 1:fdd22bb7aa52 205 blkCnt = blockSize;
emilmont 1:fdd22bb7aa52 206
emilmont 1:fdd22bb7aa52 207
mbed_official 3:7a284390b0ce 208 #ifndef ARM_MATH_CM0_FAMILY
emilmont 1:fdd22bb7aa52 209
emilmont 1:fdd22bb7aa52 210 /* Run the below code for Cortex-M4 and Cortex-M3 */
emilmont 1:fdd22bb7aa52 211
emilmont 1:fdd22bb7aa52 212 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 213 {
emilmont 1:fdd22bb7aa52 214 /* Copy the new input sample into the state buffer */
emilmont 1:fdd22bb7aa52 215 *pStateCurnt++ = *pSrc;
emilmont 1:fdd22bb7aa52 216
emilmont 1:fdd22bb7aa52 217 /* Initialize pState pointer */
emilmont 1:fdd22bb7aa52 218 px = pState;
emilmont 1:fdd22bb7aa52 219
emilmont 1:fdd22bb7aa52 220 /* Initialize coeff pointer */
emilmont 1:fdd22bb7aa52 221 pb = (pCoeffs);
emilmont 1:fdd22bb7aa52 222
emilmont 1:fdd22bb7aa52 223 /* Read the sample from input buffer */
emilmont 1:fdd22bb7aa52 224 in = *pSrc++;
emilmont 1:fdd22bb7aa52 225
emilmont 1:fdd22bb7aa52 226 /* Update the energy calculation */
emilmont 1:fdd22bb7aa52 227 energy -= x0 * x0;
emilmont 1:fdd22bb7aa52 228 energy += in * in;
emilmont 1:fdd22bb7aa52 229
emilmont 1:fdd22bb7aa52 230 /* Set the accumulator to zero */
emilmont 1:fdd22bb7aa52 231 sum = 0.0f;
emilmont 1:fdd22bb7aa52 232
emilmont 1:fdd22bb7aa52 233 /* Loop unrolling. Process 4 taps at a time. */
emilmont 1:fdd22bb7aa52 234 tapCnt = numTaps >> 2;
emilmont 1:fdd22bb7aa52 235
emilmont 1:fdd22bb7aa52 236 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 237 {
emilmont 1:fdd22bb7aa52 238 /* Perform the multiply-accumulate */
emilmont 1:fdd22bb7aa52 239 sum += (*px++) * (*pb++);
emilmont 1:fdd22bb7aa52 240 sum += (*px++) * (*pb++);
emilmont 1:fdd22bb7aa52 241 sum += (*px++) * (*pb++);
emilmont 1:fdd22bb7aa52 242 sum += (*px++) * (*pb++);
emilmont 1:fdd22bb7aa52 243
emilmont 1:fdd22bb7aa52 244 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 245 tapCnt--;
emilmont 1:fdd22bb7aa52 246 }
emilmont 1:fdd22bb7aa52 247
emilmont 1:fdd22bb7aa52 248 /* If the filter length is not a multiple of 4, compute the remaining filter taps */
emilmont 1:fdd22bb7aa52 249 tapCnt = numTaps % 0x4u;
emilmont 1:fdd22bb7aa52 250
emilmont 1:fdd22bb7aa52 251 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 252 {
emilmont 1:fdd22bb7aa52 253 /* Perform the multiply-accumulate */
emilmont 1:fdd22bb7aa52 254 sum += (*px++) * (*pb++);
emilmont 1:fdd22bb7aa52 255
emilmont 1:fdd22bb7aa52 256 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 257 tapCnt--;
emilmont 1:fdd22bb7aa52 258 }
emilmont 1:fdd22bb7aa52 259
emilmont 1:fdd22bb7aa52 260 /* The result in the accumulator, store in the destination buffer. */
emilmont 1:fdd22bb7aa52 261 *pOut++ = sum;
emilmont 1:fdd22bb7aa52 262
emilmont 1:fdd22bb7aa52 263 /* Compute and store error */
emilmont 1:fdd22bb7aa52 264 d = (float32_t) (*pRef++);
emilmont 1:fdd22bb7aa52 265 e = d - sum;
emilmont 1:fdd22bb7aa52 266 *pErr++ = e;
emilmont 1:fdd22bb7aa52 267
emilmont 1:fdd22bb7aa52 268 /* Calculation of Weighting factor for updating filter coefficients */
emilmont 1:fdd22bb7aa52 269 /* epsilon value 0.000000119209289f */
emilmont 1:fdd22bb7aa52 270 w = (e * mu) / (energy + 0.000000119209289f);
emilmont 1:fdd22bb7aa52 271
emilmont 1:fdd22bb7aa52 272 /* Initialize pState pointer */
emilmont 1:fdd22bb7aa52 273 px = pState;
emilmont 1:fdd22bb7aa52 274
emilmont 1:fdd22bb7aa52 275 /* Initialize coeff pointer */
emilmont 1:fdd22bb7aa52 276 pb = (pCoeffs);
emilmont 1:fdd22bb7aa52 277
emilmont 1:fdd22bb7aa52 278 /* Loop unrolling. Process 4 taps at a time. */
emilmont 1:fdd22bb7aa52 279 tapCnt = numTaps >> 2;
emilmont 1:fdd22bb7aa52 280
emilmont 1:fdd22bb7aa52 281 /* Update filter coefficients */
emilmont 1:fdd22bb7aa52 282 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 283 {
emilmont 1:fdd22bb7aa52 284 /* Perform the multiply-accumulate */
emilmont 1:fdd22bb7aa52 285 *pb += w * (*px++);
emilmont 1:fdd22bb7aa52 286 pb++;
emilmont 1:fdd22bb7aa52 287
emilmont 1:fdd22bb7aa52 288 *pb += w * (*px++);
emilmont 1:fdd22bb7aa52 289 pb++;
emilmont 1:fdd22bb7aa52 290
emilmont 1:fdd22bb7aa52 291 *pb += w * (*px++);
emilmont 1:fdd22bb7aa52 292 pb++;
emilmont 1:fdd22bb7aa52 293
emilmont 1:fdd22bb7aa52 294 *pb += w * (*px++);
emilmont 1:fdd22bb7aa52 295 pb++;
emilmont 1:fdd22bb7aa52 296
emilmont 1:fdd22bb7aa52 297
emilmont 1:fdd22bb7aa52 298 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 299 tapCnt--;
emilmont 1:fdd22bb7aa52 300 }
emilmont 1:fdd22bb7aa52 301
emilmont 1:fdd22bb7aa52 302 /* If the filter length is not a multiple of 4, compute the remaining filter taps */
emilmont 1:fdd22bb7aa52 303 tapCnt = numTaps % 0x4u;
emilmont 1:fdd22bb7aa52 304
emilmont 1:fdd22bb7aa52 305 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 306 {
emilmont 1:fdd22bb7aa52 307 /* Perform the multiply-accumulate */
emilmont 1:fdd22bb7aa52 308 *pb += w * (*px++);
emilmont 1:fdd22bb7aa52 309 pb++;
emilmont 1:fdd22bb7aa52 310
emilmont 1:fdd22bb7aa52 311 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 312 tapCnt--;
emilmont 1:fdd22bb7aa52 313 }
emilmont 1:fdd22bb7aa52 314
emilmont 1:fdd22bb7aa52 315 x0 = *pState;
emilmont 1:fdd22bb7aa52 316
emilmont 1:fdd22bb7aa52 317 /* Advance state pointer by 1 for the next sample */
emilmont 1:fdd22bb7aa52 318 pState = pState + 1;
emilmont 1:fdd22bb7aa52 319
emilmont 1:fdd22bb7aa52 320 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 321 blkCnt--;
emilmont 1:fdd22bb7aa52 322 }
emilmont 1:fdd22bb7aa52 323
emilmont 1:fdd22bb7aa52 324 S->energy = energy;
emilmont 1:fdd22bb7aa52 325 S->x0 = x0;
emilmont 1:fdd22bb7aa52 326
emilmont 1:fdd22bb7aa52 327 /* Processing is complete. Now copy the last numTaps - 1 samples to the
emilmont 1:fdd22bb7aa52 328 satrt of the state buffer. This prepares the state buffer for the
emilmont 1:fdd22bb7aa52 329 next function call. */
emilmont 1:fdd22bb7aa52 330
emilmont 1:fdd22bb7aa52 331 /* Points to the start of the pState buffer */
emilmont 1:fdd22bb7aa52 332 pStateCurnt = S->pState;
emilmont 1:fdd22bb7aa52 333
emilmont 1:fdd22bb7aa52 334 /* Loop unrolling for (numTaps - 1u)/4 samples copy */
emilmont 1:fdd22bb7aa52 335 tapCnt = (numTaps - 1u) >> 2u;
emilmont 1:fdd22bb7aa52 336
emilmont 1:fdd22bb7aa52 337 /* copy data */
emilmont 1:fdd22bb7aa52 338 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 339 {
emilmont 1:fdd22bb7aa52 340 *pStateCurnt++ = *pState++;
emilmont 1:fdd22bb7aa52 341 *pStateCurnt++ = *pState++;
emilmont 1:fdd22bb7aa52 342 *pStateCurnt++ = *pState++;
emilmont 1:fdd22bb7aa52 343 *pStateCurnt++ = *pState++;
emilmont 1:fdd22bb7aa52 344
emilmont 1:fdd22bb7aa52 345 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 346 tapCnt--;
emilmont 1:fdd22bb7aa52 347 }
emilmont 1:fdd22bb7aa52 348
emilmont 1:fdd22bb7aa52 349 /* Calculate remaining number of copies */
emilmont 1:fdd22bb7aa52 350 tapCnt = (numTaps - 1u) % 0x4u;
emilmont 1:fdd22bb7aa52 351
emilmont 1:fdd22bb7aa52 352 /* Copy the remaining q31_t data */
emilmont 1:fdd22bb7aa52 353 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 354 {
emilmont 1:fdd22bb7aa52 355 *pStateCurnt++ = *pState++;
emilmont 1:fdd22bb7aa52 356
emilmont 1:fdd22bb7aa52 357 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 358 tapCnt--;
emilmont 1:fdd22bb7aa52 359 }
emilmont 1:fdd22bb7aa52 360
emilmont 1:fdd22bb7aa52 361 #else
emilmont 1:fdd22bb7aa52 362
emilmont 1:fdd22bb7aa52 363 /* Run the below code for Cortex-M0 */
emilmont 1:fdd22bb7aa52 364
emilmont 1:fdd22bb7aa52 365 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 366 {
emilmont 1:fdd22bb7aa52 367 /* Copy the new input sample into the state buffer */
emilmont 1:fdd22bb7aa52 368 *pStateCurnt++ = *pSrc;
emilmont 1:fdd22bb7aa52 369
emilmont 1:fdd22bb7aa52 370 /* Initialize pState pointer */
emilmont 1:fdd22bb7aa52 371 px = pState;
emilmont 1:fdd22bb7aa52 372
emilmont 1:fdd22bb7aa52 373 /* Initialize pCoeffs pointer */
emilmont 1:fdd22bb7aa52 374 pb = pCoeffs;
emilmont 1:fdd22bb7aa52 375
emilmont 1:fdd22bb7aa52 376 /* Read the sample from input buffer */
emilmont 1:fdd22bb7aa52 377 in = *pSrc++;
emilmont 1:fdd22bb7aa52 378
emilmont 1:fdd22bb7aa52 379 /* Update the energy calculation */
emilmont 1:fdd22bb7aa52 380 energy -= x0 * x0;
emilmont 1:fdd22bb7aa52 381 energy += in * in;
emilmont 1:fdd22bb7aa52 382
emilmont 1:fdd22bb7aa52 383 /* Set the accumulator to zero */
emilmont 1:fdd22bb7aa52 384 sum = 0.0f;
emilmont 1:fdd22bb7aa52 385
emilmont 1:fdd22bb7aa52 386 /* Loop over numTaps number of values */
emilmont 1:fdd22bb7aa52 387 tapCnt = numTaps;
emilmont 1:fdd22bb7aa52 388
emilmont 1:fdd22bb7aa52 389 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 390 {
emilmont 1:fdd22bb7aa52 391 /* Perform the multiply-accumulate */
emilmont 1:fdd22bb7aa52 392 sum += (*px++) * (*pb++);
emilmont 1:fdd22bb7aa52 393
emilmont 1:fdd22bb7aa52 394 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 395 tapCnt--;
emilmont 1:fdd22bb7aa52 396 }
emilmont 1:fdd22bb7aa52 397
emilmont 1:fdd22bb7aa52 398 /* The result in the accumulator is stored in the destination buffer. */
emilmont 1:fdd22bb7aa52 399 *pOut++ = sum;
emilmont 1:fdd22bb7aa52 400
emilmont 1:fdd22bb7aa52 401 /* Compute and store error */
emilmont 1:fdd22bb7aa52 402 d = (float32_t) (*pRef++);
emilmont 1:fdd22bb7aa52 403 e = d - sum;
emilmont 1:fdd22bb7aa52 404 *pErr++ = e;
emilmont 1:fdd22bb7aa52 405
emilmont 1:fdd22bb7aa52 406 /* Calculation of Weighting factor for updating filter coefficients */
emilmont 1:fdd22bb7aa52 407 /* epsilon value 0.000000119209289f */
emilmont 1:fdd22bb7aa52 408 w = (e * mu) / (energy + 0.000000119209289f);
emilmont 1:fdd22bb7aa52 409
emilmont 1:fdd22bb7aa52 410 /* Initialize pState pointer */
emilmont 1:fdd22bb7aa52 411 px = pState;
emilmont 1:fdd22bb7aa52 412
emilmont 1:fdd22bb7aa52 413 /* Initialize pCcoeffs pointer */
emilmont 1:fdd22bb7aa52 414 pb = pCoeffs;
emilmont 1:fdd22bb7aa52 415
emilmont 1:fdd22bb7aa52 416 /* Loop over numTaps number of values */
emilmont 1:fdd22bb7aa52 417 tapCnt = numTaps;
emilmont 1:fdd22bb7aa52 418
emilmont 1:fdd22bb7aa52 419 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 420 {
emilmont 1:fdd22bb7aa52 421 /* Perform the multiply-accumulate */
emilmont 1:fdd22bb7aa52 422 *pb += w * (*px++);
emilmont 1:fdd22bb7aa52 423 pb++;
emilmont 1:fdd22bb7aa52 424
emilmont 1:fdd22bb7aa52 425 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 426 tapCnt--;
emilmont 1:fdd22bb7aa52 427 }
emilmont 1:fdd22bb7aa52 428
emilmont 1:fdd22bb7aa52 429 x0 = *pState;
emilmont 1:fdd22bb7aa52 430
emilmont 1:fdd22bb7aa52 431 /* Advance state pointer by 1 for the next sample */
emilmont 1:fdd22bb7aa52 432 pState = pState + 1;
emilmont 1:fdd22bb7aa52 433
emilmont 1:fdd22bb7aa52 434 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 435 blkCnt--;
emilmont 1:fdd22bb7aa52 436 }
emilmont 1:fdd22bb7aa52 437
emilmont 1:fdd22bb7aa52 438 S->energy = energy;
emilmont 1:fdd22bb7aa52 439 S->x0 = x0;
emilmont 1:fdd22bb7aa52 440
emilmont 1:fdd22bb7aa52 441 /* Processing is complete. Now copy the last numTaps - 1 samples to the
emilmont 1:fdd22bb7aa52 442 satrt of the state buffer. This prepares the state buffer for the
emilmont 1:fdd22bb7aa52 443 next function call. */
emilmont 1:fdd22bb7aa52 444
emilmont 1:fdd22bb7aa52 445 /* Points to the start of the pState buffer */
emilmont 1:fdd22bb7aa52 446 pStateCurnt = S->pState;
emilmont 1:fdd22bb7aa52 447
emilmont 1:fdd22bb7aa52 448 /* Copy (numTaps - 1u) samples */
emilmont 1:fdd22bb7aa52 449 tapCnt = (numTaps - 1u);
emilmont 1:fdd22bb7aa52 450
emilmont 1:fdd22bb7aa52 451 /* Copy the remaining q31_t data */
emilmont 1:fdd22bb7aa52 452 while(tapCnt > 0u)
emilmont 1:fdd22bb7aa52 453 {
emilmont 1:fdd22bb7aa52 454 *pStateCurnt++ = *pState++;
emilmont 1:fdd22bb7aa52 455
emilmont 1:fdd22bb7aa52 456 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 457 tapCnt--;
emilmont 1:fdd22bb7aa52 458 }
emilmont 1:fdd22bb7aa52 459
mbed_official 3:7a284390b0ce 460 #endif /* #ifndef ARM_MATH_CM0_FAMILY */
emilmont 1:fdd22bb7aa52 461
emilmont 1:fdd22bb7aa52 462 }
emilmont 1:fdd22bb7aa52 463
emilmont 1:fdd22bb7aa52 464 /**
emilmont 1:fdd22bb7aa52 465 * @} end of LMS_NORM group
emilmont 1:fdd22bb7aa52 466 */