progress

Makefile

@@ -1,7 +1,14 @@
+CFLAGS      += -Os -ggdb -std=gnu99 -Wfatal-errors
+CFLAGS      += -Wall -Wextra -Wshadow
+CFLAGS      += -Wwrite-strings -Wold-style-definition -Winline -Wmissing-noreturn -Wstrict-prototypes
+CFLAGS      += -Wredundant-decls -Wfloat-equal -Wsign-compare
+CFLAGS      += -Wunused-function
+CFLAGS    += -MD -lm
+
 all: main.out
 
 main.out: main.c src/vec_match.c src/vec_match.h
-	gcc -o main.out main.c src/vec_match.c -Isrc
+	gcc $(CFLAGS) -o main.out main.c src/vec_match.c -Isrc
 
 run: main.out
 	./main.out

envelope-proj.pro

@@ -4,10 +4,13 @@ CONFIG -= app_bundle
 CONFIG -= qt
 
 SOURCES += main.c \
-    vec_match.c
+    vec_match.c \
+    src/vec_match.c
 
 DISTFILES += \
-    style.astylerc
+    style.astylerc \
+    Makefile
 
 HEADERS += \
-    vec_match.h
+    vec_match.h \
+    src/vec_match.h

main.c

@@ -15,8 +15,37 @@ static float data[DATA_LEN] = {
 };
 
 
-int main()
+static float data_cprs[16] = {
+	0, 15.7, 0, 0, 0.1, 0.2, 0.1, 10, 24.242, 0, 0, 0.1, 0.2, 0.4, 0.5, 0
+};
+
+static float data_cprs2[16];
+
+int main(void)
 {
+	for (int i = 0; i < 16; i++) {
+		printf("%.1f, ", data_cprs[i]);
+	}
+	printf("\n");
+
+	int len = vec_pack(data_cprs2, 16, data_cprs, 16, 2);
+	printf("compressed len = %d\n", len);
+
+	for (int i = 0; i < len; i++) {
+		printf("%.1f, ", data_cprs2[i]);
+	}
+	printf("\n");
+
+
+	return 0;
+
+	vec_match_cfg_t cfg = {
+		.length = DATA_LEN,
+		.drift_x = 1,
+		.offset_y = 5,
+		.abs_threshold = 0.1
+	};
+
 	// example
 
 	printf("REF:  ");
@@ -36,7 +65,7 @@ int main()
 	// error metric fields
 	float env_e, abs_e;
 
-	bool ok = vec_match(data, reference, DATA_LEN, 1, 5, &env_e, &abs_e);
+	bool ok = vec_match(data, reference, &cfg, &env_e, &abs_e);
 	printf("%s", ok ? "MATCH OK" : "MATCH FAILED");
 	printf("\n");
 

main.d

@@ -0,0 +1,24 @@
+main.out: src/vec_match.c /usr/include/stdc-predef.h \
+ /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/include/stdint.h \
+ /usr/include/stdint.h /usr/include/features.h /usr/include/sys/cdefs.h \
+ /usr/include/bits/wordsize.h /usr/include/gnu/stubs.h \
+ /usr/include/gnu/stubs-64.h /usr/include/bits/wchar.h \
+ /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/include/stdbool.h \
+ /usr/include/stdlib.h \
+ /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/include/stddef.h \
+ /usr/include/bits/waitflags.h /usr/include/bits/waitstatus.h \
+ /usr/include/endian.h /usr/include/bits/endian.h \
+ /usr/include/bits/byteswap.h /usr/include/bits/types.h \
+ /usr/include/bits/typesizes.h /usr/include/bits/byteswap-16.h \
+ /usr/include/sys/types.h /usr/include/time.h /usr/include/sys/select.h \
+ /usr/include/bits/select.h /usr/include/bits/sigset.h \
+ /usr/include/bits/time.h /usr/include/sys/sysmacros.h \
+ /usr/include/bits/pthreadtypes.h /usr/include/alloca.h \
+ /usr/include/bits/stdlib-float.h \
+ /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/include/float.h \
+ /usr/include/math.h /usr/include/bits/math-vector.h \
+ /usr/include/bits/libm-simd-decl-stubs.h /usr/include/bits/huge_val.h \
+ /usr/include/bits/huge_valf.h /usr/include/bits/huge_vall.h \
+ /usr/include/bits/inf.h /usr/include/bits/nan.h \
+ /usr/include/bits/mathdef.h /usr/include/bits/mathcalls.h \
+ src/vec_match.h

src/vec_match.c

@@ -2,6 +2,7 @@
 #include <stdbool.h>
 #include <stdlib.h>
 #include <float.h>
+#include <math.h>
 
 #include "vec_match.h"
 
@@ -9,10 +10,8 @@
 
 bool vec_match(const float *data,
 			   const float *ref,
-			   uint32_t length,
-			   uint8_t drift_x,
-			   float offset_y,
-			   float *envl_match_error,
+			   const vec_match_cfg_t *cfg,
+			   float *fuzzy_match_error,
 			   float *abs_match_error)
 {
 	int a, b;
@@ -21,15 +20,15 @@ bool vec_match(const float *data,
 	float env_err = 0;
 	float abs_err = 0;
 
-	for (int i = 0; i < (int)length; i++) {
+	for (int i = 0; i < (int)cfg->length; i++) {
 		float peak = FLT_MIN;
 		float base = FLT_MAX;
 
 		// find highest value in the surrounding drift_x points
-		a = i - drift_x;
-		b = i + drift_x;
+		a = i - cfg->drift_x;
+		b = i + cfg->drift_x;
 		if (a < 0) a = 0;
-		if (b >= (int)length) b = length - 1;
+		if (b >= (int)cfg->length) b = cfg->length - 1;
 
 		for (int j = a; j <= b; j++) {
 			if (peak < ref[j]) peak = ref[j];
@@ -37,13 +36,16 @@ bool vec_match(const float *data,
 		}
 
 		// apply drift_y
-		peak += offset_y;
-		base -= offset_y;
+		peak += cfg->offset_y; // add abs threshold on top
+		base -= cfg->offset_y;
 
-		abs_err += SQUARE(ref[i] - data[i]);
+		// ignore abs threshold difference (float precision error)
+		if (fabs(ref[i] - data[i]) > cfg->abs_threshold) {
+			abs_err += SQUARE(ref[i] - data[i]);
+		}
 
 
-		if (data[i] >= base && data[i] <= peak) {
+		if (data[i] >= (base - cfg->abs_threshold) && data[i] <= (peak + cfg->abs_threshold)) {
 			// within limits
 			continue;
 		} else {
@@ -57,8 +59,54 @@ bool vec_match(const float *data,
 	}
 
 	// write error values to provided fields
-	if (envl_match_error != NULL) *envl_match_error = env_err;
+	if (fuzzy_match_error != NULL) *fuzzy_match_error = env_err;
 	if (abs_match_error != NULL) *abs_match_error = abs_err;
 
 	return err_cnt == 0;
 }
+
+
+
+
+uint32_t vec_pack(float *result, uint32_t result_capacity, const float *data, uint32_t data_length, float threshold)
+{
+	uint32_t result_len = 0;
+	uint32_t zeroes = 0;
+
+	for (uint32_t i = 0; i < data_length; i++) {
+		if (data[i] < threshold) {
+			zeroes++;
+		} else {
+			// write zero marker to result
+			if (zeroes) {
+				if (result_len < result_capacity) {
+					result[result_len] = 0.0f - zeroes; // float and negative
+				}
+
+				zeroes = 0;
+				result_len++; //length is increased even if buffer full
+			}
+
+			if (result_len < result_capacity) {
+				result[result_len] = data[i];
+			}
+
+			result_len++;
+		}
+	}
+
+	// handle trailing zeroes
+	if (zeroes) {
+		if (result_len < result_capacity) {
+			result[result_len] = 0.0f - zeroes;
+		}
+
+		result_len++;
+	}
+
+	return result_len;
+}
+
+
+
+

src/vec_match.h

@@ -3,23 +3,82 @@
 #include <stdint.h>
 #include <stdbool.h>
 
+/* Example: drift_x 1, offset_y 10, abs_threshold 0.1 */
+typedef struct {
+	uint32_t length; // data length (data & ref length must be equal)
+	uint32_t drift_x; // allowed horizontal drift (Hz drift if values are 1 Hz FFT bins)
+	float offset_y;  // allowed vertical offset (bin amplitude, positive or negative)
+	float abs_threshold; // absolute threshold (to fix preccision errors, also added to offset_y)
+} vec_match_cfg_t;
+
 
 /**
  * Match signal to reference, allowing for some offser and noise
  *
  * @param data matched data
  * @param ref reference data
- * @param length data length (data & ref length must be equal)
- * @param drift_x allowed horizontal drift (Hz drift if values are 1 Hz FFT bins)
- * @param offset_y allowed vertical offset (bin amplitude, positive or negative)
- * @param envl_match_error error metric calculated with allowed drift and offset
+ * @param cfg config struct
+ * @param fuzzy_match_error error metric calculated with allowed drift and offset
  * @param abs_match_error error metric calculated from raw data (can be used if envelope match passes)
  * @return envelope match status (match using drift and offset)
  */
 bool vec_match(const float *data,
 			   const float *ref,
-			   uint32_t length,
-			   uint8_t drift_x,
-			   float offset_y,
-			   float *envl_match_error,
+			   const vec_match_cfg_t *cfg,
+			   float *fuzzy_match_error,
 			   float *abs_match_error);
+
+
+/**
+ * Match vectors of positive numbers.
+ * Negative number indicates how many consecutive elements are zero (hence the compression).
+ *
+ * 1024-long vector [12, 0, ...] would be [12, -1023]
+ *
+ * Params otherwise the same as vec_match()
+ */
+bool vec_match_packed(const float *data,
+					const float *ref,
+					const vec_match_cfg_t *cfg,
+					float *fuzzy_match_error,
+					float *abs_match_error);
+
+
+/**
+ * Compress a vector by replacing sequence of zeroes with a negative value indicating their count.
+ *
+ * Returned length may exceed result_capacity, but the buffer is never overrun.
+ * That can be used to gradually increase the threshold until the compressed data fits in the result buffer.
+ *
+ * The compression is by definition lossy.
+ *
+ * @param result result vector (can be the same as data vector for in-place operation)
+ * @param result_capacity size of result buffer.
+ * @param data data vector
+ * @param length data legth
+ * @param threshold max value to be considered zero in the compression
+ * @return length of result vector
+ */
+uint32_t vec_pack(float *result,
+					  uint32_t result_capacity,
+					  const float *data,
+					  uint32_t length,
+					  float threshold);
+
+
+/**
+ * Unpack a vector compressed with vec_pack().
+ *
+ * If returned length exceeds provided buffer capacity, it's an indication that you need to enlarge your buffer.
+ * The buffer is never overrun, though.
+ *
+ * @param result result buffer
+ * @param result_capacity result buffer size
+ * @param compr_data compressed data vector
+ * @param compr_length compressed data vector length
+ * @return
+ */
+uint32_t vec_unpack(float *result,
+					uint32_t result_capacity,
+					const float *compr_data,
+					uint32_t compr_length);