add a stream-based analyzer interface

analyzer/analyzer.cc

@@ -35,7 +35,96 @@ using std::cout;
 using std::cerr;
 using std::endl;
 
-typedef uint16_t sample_t;
+#define STATE_PREFILL    0
+#define STATE_RUN        1
+
+StreamAnalyzer::StreamAnalyzer()
+  : hanwin(WINDOWSIZE),
+    state(STATE_PREFILL),
+    frames(0),
+    buf_pos(0),
+    outdata(NUMFREQS)
+{}
+
+void StreamAnalyzer::process(char *data, int n) {
+  char *buf_head = (char *)(indata + OVERLAP);
+  while (n > 0) {
+    int sz = sizeof(sample_t) * READSIZE - buf_pos;
+    if (n < sz)
+       sz = n;
+    memcpy(buf_head + buf_pos, data, sz);
+    n -= sz;
+    data += sz;
+    buf_pos += sz;
+    if (buf_pos == sizeof(sample_t) * READSIZE) {
+      buf_pos = 0;
+      process_buffer();
+    }
+  }
+}
+
+void StreamAnalyzer::process_buffer() {
+  switch(state) {
+  case STATE_PREFILL:
+    state = STATE_RUN;
+    break;
+  case STATE_RUN:
+    process_buffer_internal();
+  }
+
+  // shift the read data
+  memmove(indata, indata + READSIZE, OVERLAP * sizeof(sample_t));
+}
+
+void StreamAnalyzer::process_buffer_internal() {
+  if (++frames > MAXFRAMES)
+     return;
+
+  // calculate MFCCs:
+  for (int i = 0; i < WINDOWSIZE; ++i)
+     pcmfft.input()[i] = (double)indata[i];
+  
+  // window the data
+  hanwin.apply(pcmfft.input(), WINDOWSIZE);
+
+  // fft to get the spectrum
+  pcmfft.execute();
+
+  // calculate the power spectrum
+  for (int i = 0; i < NUMFREQS; ++i)
+     outdata[i] = pow(pcmfft.output()[i][0], 2) +
+        pow(pcmfft.output()[i][1], 2);
+
+  // apply mel filter bank
+  vector<double> melfreqs;
+  mfbank.apply(outdata, melfreqs);
+
+  beatkeeper.process(melfreqs);
+
+  // compute log energy
+  for (int i = 0; i < NUMMEL; ++i)
+     melfreqs[i] = log(melfreqs[i]);
+  
+  // another fft to get the MFCCs
+  specfft.apply(melfreqs);
+  
+  // discard the first mfcc
+  float cepstrum[NUMCEPSTR];
+  for (int i = 1; i <= NUMCEPSTR; ++i)
+     cepstrum[i - 1] = specfft.output()[i][0];
+  
+  mfcckeeper.process(cepstrum);
+}
+
+Features *StreamAnalyzer::get_result() {
+  if (frames < 100)
+     return NULL;
+
+  mfcckeeper.finalize();
+  beatkeeper.finalize();
+
+  return new Features(mfcckeeper.get_result(), beatkeeper.get_result());
+}
 
 // Calculate acoustic stats for a song and write them to the database.
 Features *Analyzer::analyze(int infd)

analyzer/analyzer.h

@@ -57,4 +57,31 @@ protected:
     HanningWindow hanwin;
 };
 
+typedef uint16_t sample_t;
+
+class StreamAnalyzer
+{
+public:
+   StreamAnalyzer();
+   void process(char *data, int n);
+   Features *get_result();
+protected:
+   void process_buffer();
+   void process_buffer_internal();
+
+   FFTWisdom wisdom;
+   FFTProvider<WINDOWSIZE> pcmfft;
+   FFTProvider<NUMMEL> specfft;
+   MelFilterBank mfbank;
+   HanningWindow hanwin;
+
+   int state;
+   size_t frames;
+   int buf_pos;
+   sample_t indata[WINDOWSIZE];
+   vector<double> outdata;
+   MFCCKeeper mfcckeeper;
+   BeatManager beatkeeper;
+};
+
 #endif