cat

speech signal process tools

PS_ANA(1-ESPS)                              ENTROPIC RESEARCH LAB



NNNNAAAAMMMMEEEE
     ps_ana - pitch-synchronous linear-prediction analysis

SSSSYYYYNNNNOOOOPPPPSSSSIIIISSSS
     ppppssss____aaaannnnaaaa [ ----PPPP _p_a_r_a_m__f_i_l_e ] [ ----ffff _o_u_t__f_0_f_i_l_e ] [ ----{{{{pppprrrr}}}} _r_a_n_g_e ] [
      ----ssss _r_a_n_g_e ] [ ----oooo _o_r_d_e_r ] [ ----SSSS _f_r_a_m_e__s_t_e_p ] [ ----iiii _f_r_a_m_e__s_t_e_p ]
     [ ----cccc _s_p_e_c__r_e_p ] [ ----mmmm _s_p_e_c__m_e_t_h_o_d ] [ ----wwww _w_i_n_d_o_w ] [ ----eeee _p_r_e_e_m_-
     _p_h_a_s_i_s ] [ ----zzzz ] [ ----xxxx _d_e_b_u_g__l_e_v_e_l ] iiiinnnn____ddddaaaattttaaaa iiiinnnn____ppppuuuullllsssseeee oooouuuutttt____ffffiiiilllleeee

DDDDEEEESSSSCCCCRRRRIIIIPPPPTTTTIIIIOOOONNNN
     _P_s__a_n_a performs a pitch-synchronous LP  (linear  prediction)
     analysis.   The input file _i_n__d_a_t_a is a speech signal stored
     in a FEA_SD file.  The input file _i_n__p_u_l_s_e,  typically  gen-
     erated  by  the  _e_p_o_c_h_s(_1-_E_S_P_S) program, is a FEA_SD(5-ESPS)
     file that contains pitch pulse locations of _i_n__d_a_t_a.  Either
     _i_n__d_a_t_a  or _i_n__p_u_l_s_e, but not both, may be "-", for standard
     input.

     The output file _o_u_t__f_i_l_e is a FEA file that  contains  spec-
     tral  parameters  of  the  specified type.  Analysis results
     such as signal RMS, F0, and voicing state can optionally  be
     stored  in  the  _o_u_t__f_0_f_i_l_e indicated by the -_f option.  The
     pitch-synchronously  derived  quantities  in   _o_u_t_f_i_l_e   and
     _o_u_t__f_0_f_i_l_e  are periodically resampled at 200 Hz, unless the
     frame step size is otherwise specified using the  ----SSSS  or  ----iiii
     option.  In any case the periodic resampling is performed by
     choosing the values of the pitch-synchronous parameters from
     the  glottal epoch closest to the periodic sampling instant,
     rather than by interpolation.

OOOOPPPPTTTTIIIIOOOONNNNSSSS
     The following options are supported:

     ----PPPP _p_a_r_a_m__f_i_l_e [params]
          uses the parameter  file  _p_a_r_a_m__f_i_l_e  rather  than  the
          default, which is _p_a_r_a_m_s.

     ----ffff _o_u_t__f_0_f_i_l_e
          The output file  name  of  a  file  to  store  analysis
          results  of signal RMS, F0 (fundamental frequency), and
          voicing state.  These results are  periodically  resam-
          pled at the same rate as those in the "out_file".

     ----rrrr _f_i_r_s_t:_l_a_s_t

     ----rrrr _f_i_r_s_t:+_i_n_c_r
          Determines the range of points from input file.  In the
          first form, a pair of unsigned integers gives the first
          and last points of the range.  If _f_i_r_s_t is  omitted,  1
          is  used.   If  _l_a_s_t  is omitted, the last point in the
          file is used.  The second form  is  equivalent  to  the
          first  with  _l_a_s_t  =  _f_i_r_s_t  +  _i_n_c_r.   If the range is



        Copyright Entropic Research Laboratory, Inc. 1G1        1






PS_ANA(1-ESPS)                              ENTROPIC RESEARCH LAB



          omitted, the entire file is processed.

     ----pppp    Same as the ----rrrr option.

     ----ssss _f_i_r_s_t:_l_a_s_t

     ----ssss _f_i_r_s_t:+_i_n_c_r
          Same function as the ----rrrr option, but specifies the range
          of input data in seconds.

     ----oooo _o_r_d_e_r [0]
          Order for LP analysis.  The default value 0  means  the
          program  will  choose  an order equal to _2 + (_F_s/_1_0_0_0),
          where _F_s is the sampling frequency of  the  input  data
          file _i_n__d_a_t_a.

     ----SSSS _f_r_a_m_e__s_t_e_p [_0._0_0_5 * _s_a_m_p_l_i_n_g _f_r_e_q_u_e_n_c_y]
          Specifies frame step in samples.

     ----iiii _f_r_a_m_e__s_t_e_p [_0._0_0_5]
          Specifies frame step in seconds.

     ----cccc _s_p_e_c__r_e_p [RC]
          Spectral  analysis  type.   Valid  value  is  "_R_C"  for
          reflection  coefficients,  or  "_A_F_C" for autoregressive
          filter  coefficients  (or  linear  prediction   coeffi-
          cients).

     ----mmmm _s_p_e_c__m_e_t_h_o_d [AUTOC]
          Method to compute _s_p_e_c__r_e_p.  Valid value is "_A_U_T_O_C" for
          autocorrelation method, or "_C_O_V" for covariance method.

     ----wwww _w_i_n_d_o_w [HANNING]
          Windowing  technique  used  in  the  spectral  analysis
          method.   Valid  value is "_H_A_N_N_I_N_G", "_H_A_M_M_I_N_G", "_R_E_C_T",
          or "_C_O_S_4", for Hanning, Hamming, rectangular, and  cos4
          window, respectively.

     ----pppp _p_r_e_e_m_p_h_a_s_i_s [0.95]
          First-order preemphasis filter constant used before the
          LPC analysis.  Note that the RMS values returned in the
          optional _o_u_t__f_0_f_i_l_e are based on the preemphasized sig-
          nal in the pitch-synchronous analysis window.

     ----zzzz    This option reverses the signs of output  LPC  parame-
          ters in _o_u_t_f_i_l_e.

     ----xxxx _d_e_b_u_g__l_e_v_e_l [0]
          If _d_e_b_u_g__l_e_v_e_l is  positive,  _p_s__a_n_a  prints  debugging
          messages  and  other  information on the standard error
          output.  The messages proliferate  as  the  _d_e_b_u_g__l_e_v_e_l
          increases.   If  _d_e_b_u_g__l_e_v_e_l  is  0  (the  default), no



        Copyright Entropic Research Laboratory, Inc. 2G2        2






PS_ANA(1-ESPS)                              ENTROPIC RESEARCH LAB



          messages are printed.

EEEESSSSPPPPSSSS PPPPAAAARRRRAAAAMMMMEEEETTTTEEEERRRRSSSS
     The following ESPS parameters have the same meanings as  the
     command-line options.

     _s_t_a_r_t - _i_n_t_e_g_e_r

          The first point in the input sampled data file that  is
          processed.   A  value  of 1 denotes the first sample in
          the file.  This is only read if the ----pppp  option  is  not
          used.   If it is not in the parameter file, the default
          value of 1 is used.

     _n_a_n - _i_n_t_e_g_e_r

          The total number of data points to process.  If _n_a_n  is
          0, the whole file is processed. _N_a_n is read only if the
          ----pppp option is not used.  (See  the  discussion  under  ----
          ppppffffPPPP))))....

      _f_r_a_m_e__s_t_e_p - _f_l_o_a_t
          Analysis frame step interval to use when resampling the
          output parameters.  Valid values are in the range [1/Fs
          0.1].

     _o_r_d_e_r - _i_n_t

          See OPTIONS.

     _s_p_e_c__r_e_p - _s_t_r_i_n_g

          See OPTIONS.

     _s_p_e_c__m_e_t_h_o_d - _s_t_r_i_n_g

          See OPTIONS.

     _w_i_n_d_o_w - _s_t_r_i_n_g

          See OPTIONS.

     _p_r_e_e_m_p_h_a_s_i_s - _f_l_o_a_t

          See OPTIONS.

     _p_e_r__s_t_e_p - _f_l_o_a_t

          Step size in seconds for periodic analysis in  unvoiced
          regions.

     _p_e_r__w_s_i_z_e - _f_l_o_a_t



        Copyright Entropic Research Laboratory, Inc. 3G3        3






PS_ANA(1-ESPS)                              ENTROPIC RESEARCH LAB



          Window duration in seconds  for  periodic  analysis  in
          unvoiced regions.

     _p_s__w_s_i_z_e - _f_l_o_a_t

          Window duration in fraction  of  voiced  pitch  period.
          See also the _p_h_a_s_e parameter.

     _s_t_a_b_i_l_i_t_y - _f_l_o_a_t

          Stability factor in -dB.

     _p_h_a_s_e - _f_l_o_a_t
          This is a number between 0 and 1.  Value  0  means  the
          location  of  the window center will be the same as the
          epoch pulse location.  A value of 0.75,  (the  default)
          causes  the  window to be centered in the middle of the
          "closed glottis" region, assuming an "open quotient" of
          0.5.

EEEESSSSPPPPSSSS CCCCOOOOMMMMMMMMOOOONNNN
     No ESPS common parameter processing is supported.

EEEESSSSPPPPSSSS HHHHEEEEAAAADDDDEEEERRRRSSSS
     The usual _r_e_c_o_r_d__f_r_e_q, _s_t_a_r_t__t_i_m_e header items, and all sup-
     ported ESPS parameters are stored as generic header items.

     The bandwidth of _i_n__d_a_t_a, usually one half of  its  sampling
     frequency, is recorded as a generic header item _b_a_n_d_w_i_d_t_h of
     DOUBLE type.

FFFFUUUUTTTTUUUURRRREEEE CCCCHHHHAAAANNNNGGGGEEEESSSS
EEEEXXXXAAAAMMMMPPPPLLLLEEEESSSS
     The manual page for _e_p_o_c_h_s includes an  example  of  glottal
     epoch  detection  such  as  might  be performed before using
     _p_s__a_n_a.  Assuming  one  has  a  file  "spch.pe"   containing
     impulses at the estimated instants of glottal closure, and a
     speech file called "spch.sd",  the  following  command  will
     produce  the  files  "spch.f0ps" and "spch.rc".  The ".f0ps"
     file has F0 estimates obtained by inverting the  inter-pulse
     intervals  from  the  ".pe"  file  and RMS computed from the
     epoch-synchronous windows of preemphasized data used by  the
     LPC analysis.

     ps_ana -f spch.f0ps spch.sd spch.pe spch.rc

EEEERRRRRRRROOOORRRRSSSS AAAANNNNDDDD DDDDIIIIAAAAGGGGNNNNOOOOSSSSTTTTIIIICCCCSSSS
BBBBUUUUGGGGSSSS
     None known.

RRRREEEEFFFFEEEERRRREEEENNNNCCCCEEEESSSS
     Talkin, D. and Rowley, J., "Pitch-Synchronous  analysis  and



        Copyright Entropic Research Laboratory, Inc. 4G4        4






PS_ANA(1-ESPS)                              ENTROPIC RESEARCH LAB



     synthesis for TTS systems," _P_r_o_c_e_e_d_i_n_g_s _o_f _t_h_e _E_S_C_A _W_o_r_k_s_h_o_p
     _o_n  _S_p_e_e_c_h  _S_y_n_t_h_e_s_i_s,  C.  Benoit,  Ed.,   Imprimerie   des
     Ecureuils, Gieres, France, 1990.

SSSSEEEEEEEE AAAALLLLSSSSOOOO
     lp_syn(1-ESPS), formant(1-ESPS), get_f0(1-ESPS),
     epochs(1-ESPS), get_resid(1-ESPS), FEA(5-ESPS),
     FEA_SD(5-ESPS)

AAAAUUUUTTTTHHHHOOOORRRR
     David Talkin, Derek Lin









































        Copyright Entropic Research Laboratory, Inc. 5G5        5