From 5cfa3ffe0a18d6e1338aacb1e78cea1256b986aa Mon Sep 17 00:00:00 2001
From: Andrea Eichenseer <andrea.eichenseer@iis.fraunhofer.de>
Date: Mon, 8 May 2023 13:54:21 +0200
Subject: [PATCH] Issue 440: Fix directional noise becoming diffuse in
 ParamISM, switch FIX_440_PARAM_ISM_DIR_NOISE

---
 lib_com/ivas_stat_com.h      |  3 ++
 lib_com/options.h            |  2 ++
 lib_enc/ivas_ism_param_enc.c | 70 ++++++++++++++++++++++++++++++++++++
 3 files changed, 75 insertions(+)

diff --git a/lib_com/ivas_stat_com.h b/lib_com/ivas_stat_com.h
index c1e35d822c..305a18951a 100644
--- a/lib_com/ivas_stat_com.h
+++ b/lib_com/ivas_stat_com.h
@@ -164,6 +164,9 @@ typedef struct ivas_param_ism_data_structure
 
     int16_t flag_noisy_speech;
     int16_t noisy_speech_buffer[PARAM_ISM_HYS_BUF_SIZE];
+#ifdef FIX_440_PARAM_ISM_DIR_NOISE
+    int16_t flag_equal_energy;
+#endif
 
 } PARAM_ISM_CONFIG_DATA, *PARAM_ISM_CONFIG_HANDLE;
 
diff --git a/lib_com/options.h b/lib_com/options.h
index 9fe797240d..7b91fac70f 100644
--- a/lib_com/options.h
+++ b/lib_com/options.h
@@ -177,6 +177,8 @@
 #define FIX_435_ISM_MERGE_BUG                           /* Eri: Merge bug fix for ISM NULL metadata and tcx_only cases */
 #define FIX_355_REFACTOR_PARAMBIN_TO_5MS                /* Nokia: Fixes issue 355 by refactoring parametric binauralizer code to 5 ms mode */
 
+#define FIX_440_PARAM_ISM_DIR_NOISE                     /* FhG: Issue 440: Fix directional background noise becoming diffuse in ParamISM */
+
 /* ################## End DEVELOPMENT switches ######################### */
 /* clang-format on */
 #endif
diff --git a/lib_enc/ivas_ism_param_enc.c b/lib_enc/ivas_ism_param_enc.c
index 32fb6af515..6534d215a5 100644
--- a/lib_enc/ivas_ism_param_enc.c
+++ b/lib_enc/ivas_ism_param_enc.c
@@ -55,6 +55,12 @@ static void ivas_param_ism_compute_obj_parameters(
     int16_t i, b, m, br, mr;
     int16_t brange_start, brange_end, mrange_start, mrange_end, time_merge_fac;
     float power_ratios_m[MAX_PARAM_ISM_NBANDS][MAX_PARAM_ISM_NBLOCKS];
+#ifdef FIX_440_PARAM_ISM_DIR_NOISE
+    float ref_power_local_frame[MAX_NUM_OBJECTS];
+    float tmp_ratio;
+
+    set_f( ref_power_local_frame, 0, MAX_NUM_OBJECTS );
+#endif
 
     assert( nchan_ism == 3 || nchan_ism == 4 );
 
@@ -90,6 +96,10 @@ static void ivas_param_ism_compute_obj_parameters(
                         ref_power_local[i] += reference_power_obj[i][mr][br];
                     }
                 }
+#ifdef FIX_440_PARAM_ISM_DIR_NOISE
+                /* Sum up T/F tiles per object */
+                ref_power_local_frame[i] += ref_power_local[i];
+#endif
             }
 
             /* find two dominant objects and derive object indices for current T/F tile */
@@ -145,6 +155,28 @@ static void ivas_param_ism_compute_obj_parameters(
         }
     }
 
+#ifdef FIX_440_PARAM_ISM_DIR_NOISE
+    /* Check if objects have roughly equal power by comparing reference power of first object against all others*/
+    hParamIsm->flag_equal_energy = 1;
+    for ( i = 1; i < nchan_ism; i++ )
+    {
+        if ( ref_power_local_frame[i] != 0.0f )
+        {
+            tmp_ratio = ref_power_local_frame[0] / ref_power_local_frame[i];
+
+            if ( ( tmp_ratio > 0.975f ) && ( tmp_ratio < 1.025f ) )
+            {
+                hParamIsm->flag_equal_energy &= 1;
+            }
+            else
+            {
+                hParamIsm->flag_equal_energy &= 0;
+                break;
+            }
+        }
+    }
+#endif
+
     return;
 }
 
@@ -426,6 +458,7 @@ void ivas_param_ism_compute_noisy_speech_flag(
         st_ivas->hDirAC->hParamIsm->noisy_speech_buffer[i] = st_ivas->hDirAC->hParamIsm->noisy_speech_buffer[i + 1];
     }
 
+#ifndef FIX_440_PARAM_ISM_DIR_NOISE
     /* For the current frame, make a decision based on some core-coder flags */
     if ( st_ivas->hSCE[0]->hCoreCoder[0]->flag_noisy_speech_snr && st_ivas->hSCE[1]->hCoreCoder[0]->flag_noisy_speech_snr )
     {
@@ -454,7 +487,44 @@ void ivas_param_ism_compute_noisy_speech_flag(
     {
         st_ivas->hDirAC->hParamIsm->flag_noisy_speech = st_ivas->hDirAC->hParamIsm->flag_noisy_speech && st_ivas->hDirAC->hParamIsm->noisy_speech_buffer[i];
     }
+#else
+    /* Set flag_noisy_speech to 0 for cases where object energies are not roughly equal */
+    if ( !st_ivas->hDirAC->hParamIsm->flag_equal_energy )
+    {
+        st_ivas->hDirAC->hParamIsm->noisy_speech_buffer[i] = 0;
+        st_ivas->hDirAC->hParamIsm->flag_noisy_speech = 0;
+    }
+    else
+    {
+        /* For the current frame, make a decision based on some core-coder flags */
+        if ( st_ivas->hSCE[0]->hCoreCoder[0]->flag_noisy_speech_snr && st_ivas->hSCE[1]->hCoreCoder[0]->flag_noisy_speech_snr )
+        {
+#ifdef FIX_422
+            if ( st_ivas->hSCE[0]->hCoreCoder[0]->vad_flag || st_ivas->hSCE[1]->hCoreCoder[0]->vad_flag )
+#else
+            if ( st_ivas->hSCE[0]->hCoreCoder[0]->vad_flag && st_ivas->hSCE[1]->hCoreCoder[0]->vad_flag )
+#endif
+            {
+                st_ivas->hDirAC->hParamIsm->noisy_speech_buffer[i] = 0;
+            }
+            else
+            {
+                st_ivas->hDirAC->hParamIsm->noisy_speech_buffer[i] = 1;
+            }
+        }
+        else
+        {
+            st_ivas->hDirAC->hParamIsm->noisy_speech_buffer[i] = 0;
+        }
 
+        /* Do a decision based on hysteresis */
+        st_ivas->hDirAC->hParamIsm->flag_noisy_speech = 1;
+        for ( i = 0; i < PARAM_ISM_HYS_BUF_SIZE; i++ )
+        {
+            st_ivas->hDirAC->hParamIsm->flag_noisy_speech = st_ivas->hDirAC->hParamIsm->flag_noisy_speech && st_ivas->hDirAC->hParamIsm->noisy_speech_buffer[i];
+        }
+    }
+#endif
 
     return;
 }
-- 
GitLab