diff options
Diffstat (limited to 'speech_utils')
| -rw-r--r-- | speech_utils/init.lua | 15 |
1 files changed, 10 insertions, 5 deletions
diff --git a/speech_utils/init.lua b/speech_utils/init.lua index 3d1a470..3500eb4 100644 --- a/speech_utils/init.lua +++ b/speech_utils/init.lua @@ -1,6 +1,7 @@ nerv.speech_utils = {} -function nerv.speech_utils.global_transf(feat_utter, global_transf, frm_ext, gconf) +function nerv.speech_utils.global_transf(feat_utter, global_transf, + frm_ext, frm_trim, gconf) local rearranged feat_utter = gconf.cumat_type.new_from_host(feat_utter) if frm_ext > 0 then @@ -9,8 +10,12 @@ function nerv.speech_utils.global_transf(feat_utter, global_transf, frm_ext, gco local expanded = gconf.cumat_type(feat_utter:nrow(), feat_utter:ncol() * step) expanded:expand_frm(feat_utter, frm_ext) -- rearrange the feature (``transpose'' operation in TNet) - rearranged = expanded:create() - rearranged:rearrange_frm(expanded, step) + if gconf.rearrange then + rearranged = expanded:create() + rearranged:rearrange_frm(expanded, step) + else + rearranged = expanded + end else rearranged = feat_utter end @@ -21,8 +26,8 @@ function nerv.speech_utils.global_transf(feat_utter, global_transf, frm_ext, gco global_transf:init(input[1]:nrow()) global_transf:propagate(input, output) -- trim frames - expanded = gconf.mmat_type(output[1]:nrow() - frm_ext * 2, output[1]:ncol()) - output[1]:copy_toh(expanded, frm_ext, feat_utter:nrow() - frm_ext) + expanded = gconf.mmat_type(output[1]:nrow() - frm_trim * 2, output[1]:ncol()) + output[1]:copy_toh(expanded, frm_trim, feat_utter:nrow() - frm_trim) collectgarbage("collect") return expanded end |