Merge pull request #354 from cmusphinx/339-final-state

Do not attempt to align phones to impossibly short durations (fixes #339)
cmusphinx · Jul 20, 2023 · 05cb832 · 05cb832
2 parents f899dfe + 6252f26
commit 05cb832
Show file tree

Hide file tree

Showing 2 changed files with 60 additions and 3 deletions.
diff --git a/cython/test/alignment_test.py b/cython/test/alignment_test.py
@@ -3,12 +3,13 @@
 import os
 from pocketsphinx import Decoder
 import unittest
+import wave
 
 DATADIR = os.path.join(os.path.dirname(__file__), "../../test/data")
 
 
 class TestAlignment(unittest.TestCase):
-    def _run_decode(self, decoder, expect_fail=False):
+    def _run_decode(self, decoder):
         with open(os.path.join(DATADIR, "goforward.raw"), "rb") as fh:
             buf = fh.read()
             decoder.start_utt()
@@ -45,6 +46,54 @@ def test_default_lm(self):
         self._run_decode(decoder)
         self.assertEqual(decoder.hyp().hypstr, "go forward ten meters")
 
+    def _run_phone_align(self, decoder, buf):
+        decoder.start_utt()
+        decoder.process_raw(buf, no_search=False, full_utt=True)
+        decoder.end_utt()
+        decoder.set_alignment()
+        decoder.start_utt()
+        decoder.process_raw(buf, no_search=False, full_utt=True)
+        decoder.end_utt()
+
+    def test_align_forever(self):
+        decoder = Decoder(loglevel="INFO", backtrace=True, lm=None)
+        decoder.set_align_text("feels like these days go on forever")
+        with wave.open(
+            os.path.join(DATADIR, "forever", "input_2_16k.wav"), "r"
+        ) as infh:
+            data = infh.readframes(infh.getnframes())
+            self._run_phone_align(decoder, data)
+            alignment = decoder.get_alignment()
+            phones = [entry.name for entry in alignment.phones()]
+            self.assertEqual(
+                phones,
+                [
+                    "F",
+                    "IY",
+                    "L",
+                    "Z",
+                    "L",
+                    "AY",
+                    "K",
+                    "DH",
+                    "IY",
+                    "Z",
+                    "D",
+                    "EY",
+                    "Z",
+                    "G",
+                    "OW",
+                    "AO",
+                    "N",
+                    "F",
+                    "ER",
+                    "EH",
+                    "V",
+                    "ER",
+                    "SIL",
+                ],
+            )
+
 
 if __name__ == "__main__":
     unittest.main()
diff --git a/src/state_align_search.c b/src/state_align_search.c
@@ -457,13 +457,21 @@ state_align_search_init(const char *name,
          i < sas->n_phones && itor;
          ++i, itor = ps_alignment_iter_next(itor)) {
         ps_alignment_entry_t *ent = ps_alignment_iter_get(itor);
+        int min_nframes;
+
         hmm_init(sas->hmmctx, &sas->hmms[i], FALSE,
                  ent->id.pid.ssid, ent->id.pid.tmatid);
-        if (ent-> start > 0)
+        /* Can't align less than the number of frames in an HMM! */
+        min_nframes = hmm_n_emit_state(&sas->hmms[i]);
+        if (ent->duration < min_nframes)
+            E_WARN("phone %d has impossible duration %d "
+                   "(consider disabling bestpath search)\n",
+                   i, ent->duration);
+        if (ent->start > 0 && ent->duration >= min_nframes)
             sas->sf[i] = ent->start;
         else
             sas->sf[i] = 0; /* Always active */
-        if (ent->duration > 0)
+        if (ent->duration >= min_nframes)
             sas->ef[i] = ent->start + ent->duration;
         else
             sas->ef[i] = INT_MAX; /* Always active */