diff --git a/silk/fixed/encode_frame_FIX.c b/silk/fixed/encode_frame_FIX.c
index 4f9e0861..0efa9e63 100644
--- a/silk/fixed/encode_frame_FIX.c
+++ b/silk/fixed/encode_frame_FIX.c
@@ -57,7 +57,7 @@ void silk_encode_do_VAD_FIX(
     if( psEnc->sCmn.speech_activity_Q8 < SILK_FIX_CONST( SPEECH_ACTIVITY_DTX_THRES, 8 ) ) {
         psEnc->sCmn.indices.signalType = TYPE_NO_VOICE_ACTIVITY;
         psEnc->sCmn.noSpeechCounter++;
-        if( psEnc->sCmn.noSpeechCounter < NB_SPEECH_FRAMES_BEFORE_DTX ) {
+        if( psEnc->sCmn.noSpeechCounter <= NB_SPEECH_FRAMES_BEFORE_DTX ) {
             psEnc->sCmn.inDTX = 0;
         } else if( psEnc->sCmn.noSpeechCounter > MAX_CONSECUTIVE_DTX + NB_SPEECH_FRAMES_BEFORE_DTX ) {
             psEnc->sCmn.noSpeechCounter = NB_SPEECH_FRAMES_BEFORE_DTX;
diff --git a/silk/float/encode_frame_FLP.c b/silk/float/encode_frame_FLP.c
index c3ad50a9..5db85c76 100644
--- a/silk/float/encode_frame_FLP.c
+++ b/silk/float/encode_frame_FLP.c
@@ -56,7 +56,7 @@ void silk_encode_do_VAD_FLP(
     if( psEnc->sCmn.speech_activity_Q8 < SILK_FIX_CONST( SPEECH_ACTIVITY_DTX_THRES, 8 ) ) {
         psEnc->sCmn.indices.signalType = TYPE_NO_VOICE_ACTIVITY;
         psEnc->sCmn.noSpeechCounter++;
-        if( psEnc->sCmn.noSpeechCounter < NB_SPEECH_FRAMES_BEFORE_DTX ) {
+        if( psEnc->sCmn.noSpeechCounter <= NB_SPEECH_FRAMES_BEFORE_DTX ) {
             psEnc->sCmn.inDTX = 0;
         } else if( psEnc->sCmn.noSpeechCounter > MAX_CONSECUTIVE_DTX + NB_SPEECH_FRAMES_BEFORE_DTX ) {
             psEnc->sCmn.noSpeechCounter = NB_SPEECH_FRAMES_BEFORE_DTX;