Fix errors and misleading results

sevagh · Apr 25, 2019 · 521481b · 521481b
1 parent 84815b6
commit 521481b
Show file tree

Hide file tree

Showing 20 changed files with 26 additions and 877 deletions.
diff --git a/Makefile b/Makefile
@@ -14,8 +14,7 @@ lib:
 		src/mpm.cpp \
 		src/swipe.cpp \
 		src/autocorrelation.cpp \
-		src/parabolic_interpolation.cpp \
-		src/capi.cpp
+		src/parabolic_interpolation.cpp
 
 install: lib
 	cp -r include/pitch_detection /usr/local/include

diff --git a/README.md b/README.md
@@ -1,6 +1,6 @@
 ### Pitch detection algorithms
 
-Autocorrelation-based C++ pitch detection algorithms with **O(nlogn) or lower** running time and a C API:
+Autocorrelation-based C++ pitch detection algorithms with **O(nlogn) or lower** running time:
 
 * McLeod pitch method - [2005 paper](http://miracle.otago.ac.nz/tartini/papers/A_Smarter_Way_to_Find_Pitch.pdf) - [visualization](./misc/mcleod)
 * YIN(-FFT) - [2002 paper](http://audition.ens.fr/adc/pdf/2002_JASA_YIN.pdf) - [visualization](./misc/yin)
@@ -18,24 +18,24 @@ Suggested usage of this library can be seen in the utility [wav_analyzer](./wav_
 
 All testing files are [here](./degraded_audio_tests) - the progressive degradations are described by the respective numbered JSON file, generated using [audio-degradation-toolbox](https://github.com/sevagh/audio-degradation-toolbox). The original clip is a Viola playing E3 from the [University of Iowa MIS](http://theremin.music.uiowa.edu/MIS.html).
 
-The results come from parsing the output of wav_analyzer to count how many slices of the input clip were in the ballpark of the expected value of 164.81 - I considered anything 160-169 to be acceptable:
+The results come from parsing the output of wav_analyzer to count how many 0.1s slices of the input clip were in the ballpark of the expected value of 164.81 - I considered anything 160-169 to be acceptable:
 
 | Degradation level | MPM # correct | YIN # correct | SWIPE' # correct |
 | ------------- | ------------- | ------------- | ------------- |
-| 0 | 60 | 61 | 28 |
-| 1 | 59 | 54 | 25 |
-| 2 | 51 | 46 | 23 |
-| 3 | 50 | 47 | 23 |
-| 4 | 49 | 27 | 10 |
-| 5 | 49 | 28 | 14 |
+| 0 | 26 | 22 | 5 |
+| 1 | 23 | 21 | 13 |
+| 2 | 19 | 21 | 9 |
+| 3 | 18 | 19 | 7 |
+| 4 | 19 | 19 | 6 |
+| 5 | 18 | 19 | 5 |
 
 ### Build and install
 
 Using this project should be as easy as `make && sudo make install` on Linux with a modern GCC - I don't officially support other platforms.
 
 This project depends on [ffts](https://github.com/anthonix/ffts) and BLAS/LAPACK. To run the tests, you need [googletest](https://github.com/google/googletest), and run `make -C test/ && ./test/test`. To run the bench, you need [google benchmark](https://github.com/google/benchmark), and run `make -C test/ bench && ./test/bench`.
 
-Build and install pitch_detection, run the tests, and build the examples:
+Build and install pitch_detection, run the tests, and build the sample application, wav_analyzer:
 
 ```
 # build libpitch_detection.so
@@ -51,20 +51,14 @@ make -C test clean all
 # install the library and headers to `/usr/local/lib` and `/usr/local/include`
 sudo make install
 
-# build and run C++ example
-make -C examples/cpp clean all
-./examples/cpp/example
-
-# build and run C example
-make -C examples/c clean all
-./examples/c/example
+# build and run C++ sample
+make -C wav_analyzer clean all
+./wav_analyzer/wav_analyzer
 ```
 
 ### Usage
 
-#### C++
-
-Read the [C++ header](./include/pitch_detection/pitch_detection.h) and [C++ example](./examples/cpp).
+Read the [header](./include/pitch_detection/pitch_detection.h) and [sample wav_analyzer](./wav_analyzer).
 
 The namespaces are `pitch` and `pitch_alloc`. The functions and classes are templated for `<double>` and `<float>` support.
 
@@ -77,6 +71,7 @@ std::vector<double> audio_buffer(8092);
 
 double pitch_yin = pitch::yin<double>(audio_buffer, 48000);
 double pitch_mpm = pitch::mpm<double>(audio_buffer, 48000);
+double pitch_swipe = pitch::swipe<double>(audio_buffer, 48000);
 
 std::vector<std::pair<double, double>> pitches_pyin = pitch::pyin<double>(audio_buffer, 48000);
 std::vector<std::pair<double, double>> pitches_pmpm = pitch::pmpm<double>(audio_buffer, 48000);
@@ -92,49 +87,3 @@ for (int i = 0; i < 10000; ++i) {
         auto pitches_pmpm = ma.probabilistic_pitches(audio_buffer, 48000);
 }
 ```
-
-#### C
-
-Read the [C header](./include/pitch_detection/cpitch_detection.h) and [C example](./examples/c).
-
-In C, the `pitch` and `pitch_alloc` namespaces becomes `pitch_` and `pitch_alloc_` prefixes. `double` and `float` templates are now explicit structs and functions with `d/D` and `f/F` in the name.
-
-Also, to represent `std::vector<std::pair<T, T>>` pitch candidates, there are some custom structs:
-
-```c
-struct pitch_candidates_d_t {
-        long n_candidates;
-        struct pitch_probability_pair_d_t *candidates;
-};
-
-struct pitch_probability_pair_d_t {
-        double pitch;
-        double probability;
-};
-```
-
-Here are the above C++ examples, transliterated using the C API:
-
-```c
-#include <cpitch_detection.h>
-
-double audio_buffer[8092];
-
-double pitch_yin = pitch_yin_d(audio_buffer, 8092, 48000);
-double pitch_mpm = pitch_mpm_d(audio_buffer, 8092, 48000);
-
-//pyin and pmpm emit struct
-struct pitch_candidates_d_t * pitches_pmpm = pitch_pmpm_d(audio_buffer, 8092, 48000);
-struct pitch_candidates_d_t * pitches_pyin = pitch_pyin_d(audio_buffer, 8092, 48000);
-
-struct Mpm_d_t ma = NewMpmD(8092);
-struct Yin_d_t ya = NewYinD(8092);
-
-for (int i = 0; i < 10000; ++i) {
-        double pitch_yin = ya.pitch(audio_buffer, 48000);
-        double pitch_mpm = ma.pitch(audio_buffer, 48000);
-
-        struct pitch_candidates_d_t * pitches_pmpm = pitch_alloc_pmpm_d(ma, audio_buffer, 48000);
-        struct pitch_candidates_d_t * pitches_pyin = pitch_alloc_pyin_d(ya, audio_buffer, 48000);
-}
-```
diff --git a/docs/hmm.png b/docs/hmm.png
diff --git a/docs/index.html b/docs/index.html
@@ -18,7 +18,6 @@ <h2>2 - How should I get started?</h2>
     <p>Try to write your own autocorrelation-based pitch detection algorithm, e.g. with <a href="https://docs.scipy.org/doc/scipy-0.14.0/reference/generated/scipy.signal.correlate.html">scipy</a>. Use the <a href="https://github.com/sevagh/pitch-detection/blob/master/wav_analyzer/wav_analyzer.cpp">wav_analyzer</a> utility to see what it might look like in C++ to open an audio file and send it chunk by chunk to a pitch detection algorithm. If you prefer Python, here's a tiny <a href="https://github.com/sevagh/pitch-detection/blob/master/misc/samples/get_raw_audio.py">Python snippet</a> which uses Pydub to extract audio from a file.</p>
     <h2>3 - What's next?</h2>
     <p>Read the <a href="https://www.eecs.qmul.ac.uk/~simond/pub/2014/MauchDixon-PYIN-ICASSP2014.pdf">Probabilistic YIN paper</a>, and the similar <a href="https://github.com/sevagh/pitch-detection/blob/master/misc/probabilistic-mcleod">Probabilistic McLeod pitch method doc</a>.</p>
-    <p>Explore my <a href="https://github.com/sevagh/hmm-pitch-smoothing">HMM pitch smoothing</a> code which demonstrates HMM-based pitch smoothing techniques -</p>
-    <img src="hmm.png" width="800px">
+    <p>Explore my <a href="https://github.com/sevagh/hmm-pitch-smoothing">HMM pitch smoothing</a> code which demos HMM-based pitch smoothing techniques.</p>
   </body>
 </html>
diff --git a/examples/c/.gitignore b/examples/c/.gitignore
diff --git a/examples/c/Makefile b/examples/c/Makefile
diff --git a/examples/c/example.c b/examples/c/example.c
diff --git a/examples/cpp/.gitignore b/examples/cpp/.gitignore
diff --git a/examples/cpp/Makefile b/examples/cpp/Makefile
diff --git a/examples/cpp/example.cpp b/examples/cpp/example.cpp