From 0844f28d1fcacde4f2b830ff192d501bb18eab54 Mon Sep 17 00:00:00 2001
From: Adrian Cochrane <adrian@openwork.nz>
Date: Tue, 12 Jan 2021 20:10:48 +1300
Subject: [PATCH] Initialize GStreamer for CMU PocketSphinx speech-to-text.

---
 rhapsode.cabal |  3 +--
 src/main.c     | 35 +++++++++++++++++++++++++++++++++--
 2 files changed, 34 insertions(+), 4 deletions(-)

diff --git a/rhapsode.cabal b/rhapsode.cabal
index 1dcf0cf..2e887f2 100644
--- a/rhapsode.cabal
+++ b/rhapsode.cabal
@@ -78,8 +78,7 @@ executable rhapsode
   main-is: main.c
   ghc-options: -no-hs-main -threaded
   extra-libraries: espeak-ng, sndfile
-  -- pkgconfig-depends: pocketsphinx, sphinxbase -- FIXME errors out
-  cc-options: -I/usr/include/sphinxbase -I/usr/include/pocketsphinx -I/usr/include/x86_64-linux-gnu -I/usr/include/x86_64-linux-gnu/sphinxbase -lpocketsphinx -lsphinxbase -lsphinxad
+  pkgconfig-depends: gstreamer-1.0
   build-depends:       base >=4.9 && <=4.12, rhapsode
   other-modules: Stub
   hs-source-dirs: src
diff --git a/src/main.c b/src/main.c
index f670eda..b9c657d 100644
--- a/src/main.c
+++ b/src/main.c
@@ -10,7 +10,8 @@
 #include <espeak-ng/espeak_ng.h>
 #include <sndfile.h>
 
-#include <pocketsphinx/pocketsphinx.h>
+#include <gst/gst.h>
+#include <glib.h>
 
 // #define WITH_SPEECHD // FIXME Doesn't support audio cues, navigation, or even read the full page.
 #ifdef WITH_SPEECHD
@@ -341,6 +342,8 @@ FILE *parse_opt_file() {
 int main(int argc, char **argv) {
     int speak_err = 0;
     hs_init(&argc, &argv);
+    gst_init(&argc, &argv);
+    GMainLoop *loop = g_main_loop_new(NULL, FALSE);
 
     char *mimes = "text/html text/xml application/xml application/xhtml+xml text/plain";
     char *logpath = NULL;
@@ -351,11 +354,12 @@ int main(int argc, char **argv) {
     #ifdef WITH_SPEECHD
     SPDConnection *spd_conn = NULL;
     #endif
+    GstElement *stt_pipeline = NULL;
 
     int c;
     opterr = 0;
     #ifdef WITH_SPEECHD
-    while ((c = getopt(argc, argv, "xs::l::L:kw::dh")) != -1) {
+    while ((c = getopt(argc, argv, "xs::l::L:kw::dvh")) != -1) {
     #else
     while ((c = getopt(argc, argv, "xs::l::kw::h")) != -1) {
     #endif
@@ -393,6 +397,33 @@ int main(int argc, char **argv) {
             spd_set_data_mode(spd_conn, SPD_DATA_SSML);
             break;
         #endif
+        case 'v':
+            stt_pipeline = gst_pipeline_new("speech2text");
+            GstElement *src = gst_element_factory_make("autoaudiosrc", "microphone");
+            GstElement *convert = gst_element_factory_make("audioconvert", "convert");
+            GstElement *resample = gst_element_factory_make("audioresample", "resample");
+            GstElement *decoder = gst_element_factory_make("pocketsphinx", "asr");
+            GstElement *sink = gst_element_factory_make("fakesink", "output");
+
+            if (!src || !convert || !resample || !decoder || !sink) {
+                fprintf(stderr, "Failed to initialize voice recognition. You'll just have to type your commands.\n");
+                gst_object_unref(stt_pipeline);
+                stt_pipeline = NULL;
+                break;
+            }
+
+            // FIXME: Set something more appropriate
+            g_object_set(G_OBJECT(decoder), "lmctl", "test.lmctl", NULL);
+            g_object_set(G_OBJECT(decoder), "lmname", "tidigits", NULL);
+
+            // FIXME: Implement callback function
+            /*GstBus *bus = gst_pipeline_get_bus(GST_PIPELINE(pipeline));
+            bus_watch_id = gst_bus_add_watch(bus, stt_bus_call, loop); // FIXME clean this up on shutdown.
+            gst_object_unref(bus);*/
+
+            gst_bin_add_many(GST_BIN(stt_pipeline), src, convert, resample, decoder, sink, NULL);
+            gst_element_link_many(src, convert, resample, decoder, sink, NULL);
+            break;
         case '?':
             fprintf(stderr, "Invalid flag %c\n\n", optopt);
         case 'h':
-- 
2.30.2