1 | package org.apollo.meldex;
|
---|
2 |
|
---|
3 | import java.io.ByteArrayInputStream;
|
---|
4 | import java.io.ByteArrayOutputStream;
|
---|
5 | import java.io.File;
|
---|
6 | import java.io.IOException;
|
---|
7 |
|
---|
8 | import javax.sound.sampled.AudioFormat;
|
---|
9 | import javax.sound.sampled.AudioInputStream;
|
---|
10 | import javax.sound.sampled.AudioSystem;
|
---|
11 | import javax.sound.sampled.DataLine;
|
---|
12 | import javax.sound.sampled.TargetDataLine;
|
---|
13 |
|
---|
14 | /**
|
---|
15 | * Calculates a "Score" on how well two given
|
---|
16 | *
|
---|
17 | */
|
---|
18 | class MeldexMatch {
|
---|
19 |
|
---|
20 | final protected int MIC_FREQUENCY = 8;
|
---|
21 |
|
---|
22 | protected DataLine.Info micInfo_;
|
---|
23 |
|
---|
24 | protected AudioFormat micFormat_;
|
---|
25 |
|
---|
26 | protected TargetDataLine micLine_;
|
---|
27 |
|
---|
28 | protected Transcriber transcriber_ = null;
|
---|
29 |
|
---|
30 | public MeldexMatch() {
|
---|
31 | transcriber_ = new Transcriber();
|
---|
32 | }
|
---|
33 |
|
---|
34 | protected void establishMicInfo() {
|
---|
35 | AudioFormat micFormat22050, micFormat11025, micFormat8000, micFormat8096;
|
---|
36 | DataLine.Info micInfo22050, micInfo11025, micInfo8000, micInfo8096;
|
---|
37 |
|
---|
38 | // Define the format for our microphone data line
|
---|
39 | AudioFormat.Encoding encoding = AudioFormat.Encoding.PCM_UNSIGNED;
|
---|
40 | int sampleSize = 8;
|
---|
41 | int channels = 1;
|
---|
42 | boolean bigEndian = false;
|
---|
43 | float rate;
|
---|
44 |
|
---|
45 | // Define a variety of formats, each with a different sampling rate
|
---|
46 | rate = 22050F;
|
---|
47 | micFormat22050 = new AudioFormat(encoding, rate, sampleSize, channels,
|
---|
48 | (sampleSize / 8) * channels, rate, bigEndian);
|
---|
49 | rate = 11025F;
|
---|
50 | micFormat11025 = new AudioFormat(encoding, rate, sampleSize, channels,
|
---|
51 | (sampleSize / 8) * channels, rate, bigEndian);
|
---|
52 | rate = 8000F;
|
---|
53 | micFormat8000 = new AudioFormat(encoding, rate, sampleSize, channels,
|
---|
54 | (sampleSize / 8) * channels, rate, bigEndian);
|
---|
55 | rate = 8096F;
|
---|
56 | micFormat8096 = new AudioFormat(encoding, rate, sampleSize, channels,
|
---|
57 | (sampleSize / 8) * channels, rate, bigEndian);
|
---|
58 |
|
---|
59 | // Describe a variety of data lines, each with a different sampling rate
|
---|
60 | micInfo22050 = new DataLine.Info(TargetDataLine.class, micFormat22050);
|
---|
61 | micInfo11025 = new DataLine.Info(TargetDataLine.class, micFormat11025);
|
---|
62 | micInfo8000 = new DataLine.Info(TargetDataLine.class, micFormat8000);
|
---|
63 | micInfo8096 = new DataLine.Info(TargetDataLine.class, micFormat8096);
|
---|
64 |
|
---|
65 | // Try to find a microphone line that we can use
|
---|
66 | if (AudioSystem.isLineSupported(micInfo22050)) {
|
---|
67 | micInfo_ = micInfo22050;
|
---|
68 | micFormat_ = micFormat22050;
|
---|
69 | } else if (AudioSystem.isLineSupported(micInfo11025)) {
|
---|
70 | micInfo_ = micInfo11025;
|
---|
71 | micFormat_ = micFormat11025;
|
---|
72 | } else if (AudioSystem.isLineSupported(micInfo8000)) {
|
---|
73 | micInfo_ = micInfo8000;
|
---|
74 | micFormat_ = micFormat8000;
|
---|
75 | } else if (AudioSystem.isLineSupported(micInfo8096)) {
|
---|
76 | micInfo_ = micInfo8096;
|
---|
77 | micFormat_ = micFormat8096;
|
---|
78 | } else {
|
---|
79 | micInfo_ = null;
|
---|
80 | micFormat_ = null;
|
---|
81 | }
|
---|
82 | }
|
---|
83 |
|
---|
84 | protected boolean initMic() {
|
---|
85 | establishMicInfo();
|
---|
86 |
|
---|
87 | // Get the microphone line
|
---|
88 | try {
|
---|
89 | micLine_ = (TargetDataLine) AudioSystem.getLine(micInfo_);
|
---|
90 | } catch (Exception ex) {
|
---|
91 | System.err
|
---|
92 | .println("Exception occurred getting the microphone line.\n\n"
|
---|
93 | + ex);
|
---|
94 | return false;
|
---|
95 | }
|
---|
96 |
|
---|
97 | // Calculate the size of the microphone buffer required
|
---|
98 | int frameSize = (micFormat_.getChannels() * (micFormat_
|
---|
99 | .getSampleSizeInBits() / 8));
|
---|
100 | int bytesPerSecond = (int) (micFormat_.getSampleRate() * frameSize);
|
---|
101 | int micBufferSize = (int) (bytesPerSecond / MIC_FREQUENCY);
|
---|
102 |
|
---|
103 | // Open the microphone data line for recording
|
---|
104 | try {
|
---|
105 | micLine_.open(micFormat_, micBufferSize);
|
---|
106 | } catch (Exception ex) {
|
---|
107 | System.err
|
---|
108 | .println("Exception occurred opening the microphone line.\n\n"
|
---|
109 | + ex);
|
---|
110 | return false;
|
---|
111 | }
|
---|
112 |
|
---|
113 | return true;
|
---|
114 | }
|
---|
115 |
|
---|
116 | protected void record_prompt(int p, int fixed) {
|
---|
117 | System.out.print("Recording: [");
|
---|
118 | for (int i = 0; i < p; i++) {
|
---|
119 | System.out.print(".");
|
---|
120 | }
|
---|
121 | for (int i = p; i < fixed; i++) {
|
---|
122 | System.out.print(" ");
|
---|
123 | }
|
---|
124 |
|
---|
125 | System.out.print("]\r");
|
---|
126 | System.out.flush();
|
---|
127 | }
|
---|
128 |
|
---|
129 | public AudioInputStream record() {
|
---|
130 | // Make sure that we found a supported microphone line
|
---|
131 | if (micInfo_ == null) {
|
---|
132 | System.err
|
---|
133 | .println("Sorry, a suitable microphone line could not be found.");
|
---|
134 | return null;
|
---|
135 | }
|
---|
136 |
|
---|
137 | // Create a buffer for the recorded data
|
---|
138 | byte[] micBuffer = new byte[micLine_.getBufferSize()];
|
---|
139 |
|
---|
140 | // Create an output stream to save the recorded data
|
---|
141 | ByteArrayOutputStream micOut = new ByteArrayOutputStream();
|
---|
142 |
|
---|
143 | // Start the microphone actually recording
|
---|
144 | micLine_.start();
|
---|
145 |
|
---|
146 | // Continue recording until the Stop button is pressed
|
---|
147 | int fixed = 30;
|
---|
148 | int i = 0;
|
---|
149 | while (i <= fixed) {
|
---|
150 | record_prompt(i, fixed);
|
---|
151 |
|
---|
152 | // Read some data from the microphone line...
|
---|
153 | int numBytesRead = micLine_.read(micBuffer, 0, micLine_
|
---|
154 | .getBufferSize());
|
---|
155 |
|
---|
156 | // ...save it to the output stream
|
---|
157 | micOut.write(micBuffer, 0, numBytesRead);
|
---|
158 |
|
---|
159 | i++;
|
---|
160 | }
|
---|
161 | System.out.println("");
|
---|
162 |
|
---|
163 | // We have finished recording, so stop and close the microphone line
|
---|
164 | micLine_.stop();
|
---|
165 | micLine_.close();
|
---|
166 |
|
---|
167 | // Flush and close the output stream
|
---|
168 | try {
|
---|
169 | micOut.flush();
|
---|
170 | micOut.close();
|
---|
171 | } catch (Exception ex) {
|
---|
172 | System.err
|
---|
173 | .println("Exception occurred flushing/closing output stream.\n\n"
|
---|
174 | + ex);
|
---|
175 | return null;
|
---|
176 | }
|
---|
177 |
|
---|
178 | // Load the data recorded into the audio input stream ready for playback/processing
|
---|
179 | byte audioBytes[] = micOut.toByteArray();
|
---|
180 | ByteArrayInputStream bais = new ByteArrayInputStream(audioBytes);
|
---|
181 | AudioInputStream audioData = new AudioInputStream(bais, micFormat_,
|
---|
182 | audioBytes.length / micFormat_.getFrameSize());
|
---|
183 |
|
---|
184 | return audioData;
|
---|
185 | }
|
---|
186 |
|
---|
187 | public void doMatch(WavSample sample_query, WavSample sample_track) throws IOException {
|
---|
188 |
|
---|
189 | RogTrack transcribed_query = transcriber_
|
---|
190 | .transcribeSample(sample_query);
|
---|
191 |
|
---|
192 | RogTrack transcribed_track = transcriber_
|
---|
193 | .transcribeSample(sample_track);
|
---|
194 |
|
---|
195 | System.out.println("Query");
|
---|
196 | System.out.println(transcribed_query);
|
---|
197 | System.out.println("Track");
|
---|
198 | System.out.println(transcribed_track);
|
---|
199 |
|
---|
200 | Melody melody_query = transcribed_query.toMelody();
|
---|
201 | Melody melody_track = transcribed_track.toMelody();
|
---|
202 |
|
---|
203 | DynamicProgrammingAlgorithm dpa = new McNabMongeauSankoffAlgorithm();
|
---|
204 | // int match_mode = DynamicProgrammingAlgorithm.MATCH_AT_START;
|
---|
205 | int match_mode = DynamicProgrammingAlgorithm.MATCH_ANYWHERE;
|
---|
206 |
|
---|
207 | float score = dpa
|
---|
208 | .matchToPattern(melody_query, melody_track, match_mode);
|
---|
209 |
|
---|
210 | System.out.println("Edit distance = " + score);
|
---|
211 |
|
---|
212 | }
|
---|
213 |
|
---|
214 | public void liveQuery(String audiotrack) {
|
---|
215 | File input_file = new File(audiotrack);
|
---|
216 |
|
---|
217 | WavSample sample_track = new WavSample();
|
---|
218 |
|
---|
219 | if (sample_track.loadFromFile(input_file) != true) {
|
---|
220 | System.err.println("Error: Unable to load wav file: " + audiotrack);
|
---|
221 | } else {
|
---|
222 | initMic();
|
---|
223 | AudioInputStream audioData = record();
|
---|
224 |
|
---|
225 | WavSample sample_query = new WavSample(audioData);
|
---|
226 |
|
---|
227 | try {
|
---|
228 | doMatch(sample_query, sample_track);
|
---|
229 | } catch (IOException e) {
|
---|
230 | // TODO Auto-generated catch block
|
---|
231 | e.printStackTrace();
|
---|
232 | }
|
---|
233 | }
|
---|
234 |
|
---|
235 | }
|
---|
236 |
|
---|
237 | public void cannedQuery(String audioquery, String audiotrack) {
|
---|
238 | File file_query = new File(audioquery);
|
---|
239 | File file_track = new File(audiotrack);
|
---|
240 |
|
---|
241 | WavSample sample_query = new WavSample();
|
---|
242 | WavSample sample_track = new WavSample();
|
---|
243 |
|
---|
244 | if (sample_query.loadFromFile(file_query) != true) {
|
---|
245 | System.err.println("Error: Unable to load wav file: " + audioquery);
|
---|
246 | } else if (sample_track.loadFromFile(file_track) != true) {
|
---|
247 | System.err.println("Error: Unable to load wav file: " + audiotrack);
|
---|
248 | } else {
|
---|
249 | try {
|
---|
250 | doMatch(sample_query, sample_track);
|
---|
251 | } catch (IOException e) {
|
---|
252 | // TODO Auto-generated catch block
|
---|
253 | e.printStackTrace();
|
---|
254 | }
|
---|
255 | }
|
---|
256 | }
|
---|
257 |
|
---|
258 | public static void main(String args[]) {
|
---|
259 | MeldexMatch meldex = new MeldexMatch();
|
---|
260 |
|
---|
261 | int argc = args.length;
|
---|
262 | if (argc == 1) {
|
---|
263 | meldex.liveQuery(args[0]);
|
---|
264 | } else if (argc == 2) {
|
---|
265 | meldex.cannedQuery(args[0], args[1]);
|
---|
266 | } else {
|
---|
267 | System.err.println("Usage: MeldexMatch wav-file1 [wav-file2]");
|
---|
268 | System.exit(-1);
|
---|
269 | }
|
---|
270 | }
|
---|
271 | }
|
---|