summaryrefslogblamecommitdiff
path: root/init.lua
blob: fec72090410415e08bc7b5465e10a742e45d3088 (plain) (tree)
1
2
3
4
5
6
7
8





                                                                   

                                                  



                                  










                                                                


                              


                                   


                                


                                                           

                                                                                       
                                                              

                                            
                         

                                        
                
                                            
                                               
                  

                                                                                     
                                
                               
                                           


                                                        

                           
                               
                         
                             

              
require 'libspeech'
local TNetReader = nerv.class("nerv.TNetReader", "nerv.DataReader")

function TNetReader:__init(global_conf, reader_conf)
    self.feat_id = reader_conf.id
    self.frm_ext = reader_conf.frm_ext
    self.gconf = global_conf
    self.global_transf = reader_conf.global_transf
    self.debug = global_conf.debug
    if self.debug == nil then
        self.debug = false
    end
    self.feat_repo = nerv.TNetFeatureRepo(reader_conf.scp_file,
                                          reader_conf.conf_file,
                                          reader_conf.frm_ext)
    self.lab_repo = {}
    for id, mlf_spec in pairs(reader_conf.mlfs) do
        self.lab_repo[id] = nerv.TNetLabelRepo(mlf_spec.file,
                                            mlf_spec.format,
                                            mlf_spec.format_arg,
                                            mlf_spec.dir,
                                            mlf_spec.ext)
    end
end

function TNetReader:get_data()
    if self.feat_repo:is_end() then
        return nil
    end
    local res = {}
    local frm_ext = self.frm_ext
    local step = frm_ext * 2 + 1
    -- read HTK feature
    local feat_utter = self.feat_repo:cur_utter(self.debug)
    -- expand the feature
    local expanded = self.gconf.cumat_type(feat_utter:nrow(), feat_utter:ncol() * step)
    expanded:expand_frm(self.gconf.cumat_type.new_from_host(feat_utter), frm_ext)
    -- rearrange the feature (``transpose'' operation in TNet)
    local rearranged = expanded:create()
    rearranged:rearrange_frm(expanded, step)
    -- prepare for transf
    local input = {rearranged}
    local output = {rearranged:create()}
    -- do transf
    self.global_transf:init(input[1]:nrow())
    self.global_transf:propagate(input, output)
    -- trim frames
    expanded = self.gconf.mmat_type(output[1]:nrow() - frm_ext * 2, output[1]:ncol())
    output[1]:copy_toh(expanded, frm_ext, feat_utter:nrow() - frm_ext)
    res[self.feat_id] = expanded
    -- add corresponding labels
    for id, repo in pairs(self.lab_repo) do
        local lab_utter = repo:get_utter(self.feat_repo,
                                        expanded:nrow(),
                                        self.debug)
        res[id] = lab_utter
    end
    -- move the pointer to next
    self.feat_repo:next()
    collectgarbage("collect")
    return res
end