Commit 63f7d77a authored by malenovsky's avatar malenovsky
Browse files

Merge branch '412-s-m-misclassifications-on-deep-voices' into 'main' 

[Non-BE] Resolve "S/M misclassifications on deep voices"

See merge request !560
parents 82b4f423 bab6775d

lib_com/cnst.h

+4 −1
Original line number Diff line number Diff line
@@ -1532,7 +1532,10 @@ typedef enum _DCTTYPE 
#define TOD_NSPEC                           80                      /* number of spectral bins of the tonal detector */

#define TOD_THR_MASS                        0.86f                   /* initial value for the adaptive threshold of the tonal detector */

#define P2A_FACT                            0.9f                    /* long-term averaging factor for peak-to-average ratio */

#define THR_P2A                             80.0f                   /* threshold to detect strongly peaky signals */

#ifdef FIX_SP2A

#define THR_P2A_HIGH                        95.0f                   /* higher threshold to detect strongly peaky signals at low bitrates*/

#endif

#define THR_P2A                             80.0f                   /* lower threshold to detect strongly peaky signals at higher bitrates */



/*----------------------------------------------------------------------------------*

 * LD music post-filter constants

lib_com/options.h

+1 −0
Original line number Diff line number Diff line
@@ -162,6 +162,7 @@ 
#define FIX_406_IVAS_POSITION                           /* Eri: Issue 406: Unify IVAS_POSITION to use IVAS_VECTOR3 instead */

#define REND_DEBUGGING_REVISION                         /* VA: encapsulate rendering debugging options with DEBUGGING */

#define FIX_416_ISM_BR_SWITCHING                        /* FhG: add missing CLDFB reconfig to ISM BR switching */

#define FIX_SP2A                                        /* VA: Issue 412: Adjust threshold for the S_p2a feature in the tonal detector */

lib_enc/speech_music_classif.c

+18 −0
Original line number Diff line number Diff line
@@ -1828,6 +1828,9 @@ void ivas_smc_mode_selection( 
    float ton;

    int16_t i;

    float S_p2a, S_max, S_ave;

#ifdef FIX_SP2A

    float thr_sp2a;

#endif



    SP_MUS_CLAS_HANDLE hSpMusClas = st->hSpMusClas;
@@ -1858,8 +1861,23 @@ void ivas_smc_mode_selection( 
    S_ave = sum_f( st->hSpMusClas->tod_lt_Bin_E, TOD_NSPEC ) / TOD_NSPEC;

    S_p2a = S_max - S_ave;



#ifdef FIX_SP2A

    if ( element_brate <= IVAS_16k4 )

    {

        thr_sp2a = THR_P2A_HIGH;

    }

    else

    {

        thr_sp2a = THR_P2A;

    }

#endif



    /* initial 3-way selection of coding modes (ACELP/GSC/TCX) */

#ifdef FIX_SP2A

    if ( relE > -10.0f && ( S_p2a > thr_sp2a || ton > hSpMusClas->tod_thr_lt ) )

#else

    if ( relE > -10.0f && ( S_p2a > THR_P2A || ton > hSpMusClas->tod_thr_lt ) )

#endif

    {

        /* select TCX to encode extremely peaky signals or strongly tonal signals */

        st->sp_aud_decision1 = 1;