cosmetic changes

README

@@ -30,13 +30,15 @@ KALDI_ROOT=/path/of/kaldi-trunk make
 
 This should result in 'libgstkaldionline2.so'.
 
-Test id GStreamer can access the plugin:
+Test if GStreamer can access the plugin:
 
 GST_PLUGIN_PATH=. gst-inspect-1.0 kaldinnet2onlinedecoder
 
 
-
 HOW TO USE IT
 -------------
 
-TODO
+Command-line usage is demonstrated in `demo/`.
+
+Usage through GSTreamer's Python bindings is demonstrated in 
+https://github.com/alumae/kaldi-gstreamer-server (kaldigstserver/decoder2.py).

demo/prepare-models.sh

+#! /bin/bash
+
+BASE_URL=http://kaldi-asr.org/downloads/build/2/sandbox/online/egs/fisher_english/s5
+
+MODEL=exp/nnet2_online/nnet_a_gpu_online
+GRAPH=exp/tri5a
+
+wget -N $BASE_URL/$MODEL/final.mdl || exit 1
+(mkdir -p ivector_extractor; cd ivector_extractor; wget -N $BASE_URL/$MODEL/ivector_extractor/{final.ie,final.dubm,final.mat,global_cmvn.stats}) || exit 1
+(mkdir -p conf; cd conf; wget -N $BASE_URL/$MODEL/conf/{ivector_extractor.conf,online_nnet2_decoding.conf,mfcc.conf,online_cmvn.conf,splice.conf}) || exit 1
+
+wget -N $BASE_URL/$GRAPH/graph/HCLG.fst || exit 1
+wget -N $BASE_URL/$GRAPH/graph/words.txt || exit 1
+
+
+cat conf/ivector_extractor.conf | perl -npe 's/=.*nnet_a_gpu_online\//=/' > conf/ivector_extractor.fixed.conf

demo/transcribe-audio.sh

+#!/bin/bash
+
+if [ $# != 1 ]; then
+    echo "Usage: transcribe-audio.sh <audio>"
+    echo "e.g.: transcribe-audio.sh dr_strangelove.mp3" 
+    exit 1;
+fi
+
+! GST_PLUGIN_PATH=../src gst-inspect-1.0 kaldinnet2onlinedecoder > /dev/null 2>&1 && echo "Compile the plugin in ../src first" && exit 1;
+
+if [ ! -f HCLG.fst ]; then
+    echo "Run ./prepare-models.sh first to download models"
+    exit 1;
+fi
+
+audio=$1
+
+GST_PLUGIN_PATH=../src gst-launch-1.0 --gst-debug="" -q filesrc location=$audio ! decodebin ! audioconvert ! audioresample ! \
+kaldinnet2onlinedecoder \
+  model=final.mdl \
+  fst=HCLG.fst \
+  word-syms=words.txt \
+  feature-type=mfcc \
+  mfcc-config=conf/mfcc.conf \
+  ivector-extraction-config=conf/ivector_extractor.fixed.conf \
+  max-active=7000 \
+  beam=11.0 \
+  lattice-beam=5.0 \
+  do-endpointing=true \
+  endpoint-silence-phones="1:2:3:4:5:6:7:8:9:10" \
+! filesink location=/dev/stdout buffer-mode=2