Android/frameworks_base
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Apply EQ on the output of the speech synthesis only when using the Pico

Browse Source 
TTS engine.
This commit is contained in:
Jean-Michel Trivi
2010-01-11 14:04:50 -08:00
parent 9b0b155e60
commit 0320f8bdfb
2 changed files with 76 additions and 22 deletions
Download Patch File Download Diff File Expand all files Collapse all files

78
packages/TtsService/jni/android_tts_SynthProxy.cpp
View File

@@ -41,8 +41,6 @@
#define FILTER_TRANSITION_FREQ 1100.0f // in Hz
#define FILTER_SHELF_SLOPE 1.0f // Q
#define FILTER_GAIN 5.5f // linear gain
// such a huge gain is justified by how much energy in the low frequencies is "wasted" at the output
// of the synthesis. The low shelving filter removes it, leaving room for amplification.
#define USAGEMODE_PLAY_IMMEDIATELY 0
#define USAGEMODE_WRITE_TO_FILE 1
@@ -80,13 +78,19 @@ double out0;// y[n]
double out1;// y[n-1]
double out2;// y[n-2]
static float fFilterLowshelfAttenuation = FILTER_LOWSHELF_ATTENUATION;
static float fFilterTransitionFreq = FILTER_TRANSITION_FREQ;
static float fFilterShelfSlope = FILTER_SHELF_SLOPE;
static float fFilterGain = FILTER_GAIN;
static bool bUseFilter = false;
void initializeEQ() {
amp = float(pow(10.0, FILTER_LOWSHELF_ATTENUATION / 40.0));
w = 2.0 * M_PI * (FILTER_TRANSITION_FREQ / DEFAULT_TTS_RATE);
amp = float(pow(10.0, fFilterLowshelfAttenuation / 40.0));
w = 2.0 * M_PI * (fFilterTransitionFreq / DEFAULT_TTS_RATE);
sinw = float(sin(w));
cosw = float(cos(w));
beta = float(sqrt(amp)/FILTER_SHELF_SLOPE);
beta = float(sqrt(amp)/fFilterShelfSlope);
// initialize low-shelf parameters
b0 = amp * ((amp+1.0F) - ((amp-1.0F)*cosw) + (beta*sinw));
@@ -96,9 +100,9 @@ void initializeEQ() {
a1 = 2.0F * ((amp-1.0F) + ((amp+1.0F)*cosw));
a2 = -((amp+1.0F) + ((amp-1.0F)*cosw) - (beta*sinw));
m_fa = FILTER_GAIN * b0/a0;
m_fb = FILTER_GAIN * b1/a0;
m_fc = FILTER_GAIN * b2/a0;
m_fa = fFilterGain * b0/a0;
m_fb = fFilterGain * b1/a0;
m_fc = fFilterGain * b2/a0;
m_fd = a1/a0;
m_fe = a2/a0;
}
@@ -284,7 +288,9 @@ static tts_callback_status ttsSynthDoneCB(void *& userdata, uint32_t rate,
if (bufferSize > 0) {
prepAudioTrack(pJniData, pForAfter->streamType, rate, (AudioSystem::audio_format)format, channel);
if (pJniData->mAudioOut) {
applyFilter((int16_t*)wav, bufferSize/2);
if (bUseFilter) {
applyFilter((int16_t*)wav, bufferSize/2);
}
pJniData->mAudioOut->write(wav, bufferSize);
memset(wav, 0, bufferSize);
//LOGV("AudioTrack wrote: %d bytes", bufferSize);
@@ -300,7 +306,9 @@ static tts_callback_status ttsSynthDoneCB(void *& userdata, uint32_t rate,
return TTS_CALLBACK_HALT;
}
if (bufferSize > 0){
applyFilter((int16_t*)wav, bufferSize/2);
if (bUseFilter) {
applyFilter((int16_t*)wav, bufferSize/2);
}
fwrite(wav, 1, bufferSize, pForAfter->outputFile);
memset(wav, 0, bufferSize);
}
@@ -334,23 +342,50 @@ static tts_callback_status ttsSynthDoneCB(void *& userdata, uint32_t rate,
// ----------------------------------------------------------------------------
static void
static int
android_tts_SynthProxy_setLowShelf(JNIEnv *env, jobject thiz, jboolean applyFilter,
jfloat filterGain, jfloat attenuationInDb, jfloat freqInHz, jfloat slope)
{
int result = TTS_SUCCESS;
bUseFilter = applyFilter;
if (applyFilter) {
fFilterLowshelfAttenuation = attenuationInDb;
fFilterTransitionFreq = freqInHz;
fFilterShelfSlope = slope;
fFilterGain = filterGain;
if (fFilterShelfSlope != 0.0f) {
initializeEQ();
} else {
LOGE("Invalid slope, can't be null");
result = TTS_FAILURE;
}
}
return result;
}
// ----------------------------------------------------------------------------
static int
android_tts_SynthProxy_native_setup(JNIEnv *env, jobject thiz,
jobject weak_this, jstring nativeSoLib)
{
int result = TTS_FAILURE;
bUseFilter = false;
SynthProxyJniStorage* pJniStorage = new SynthProxyJniStorage();
prepAudioTrack(pJniStorage,
DEFAULT_TTS_STREAM_TYPE, DEFAULT_TTS_RATE, DEFAULT_TTS_FORMAT, DEFAULT_TTS_NB_CHANNELS);
const char *nativeSoLibNativeString =
env->GetStringUTFChars(nativeSoLib, 0);
const char *nativeSoLibNativeString = env->GetStringUTFChars(nativeSoLib, 0);
void *engine_lib_handle = dlopen(nativeSoLibNativeString,
RTLD_NOW | RTLD_LOCAL);
if (engine_lib_handle == NULL) {
LOGE("android_tts_SynthProxy_native_setup(): engine_lib_handle == NULL");
// TODO report error so the TTS can't be used
} else {
TtsEngine *(*get_TtsEngine)() =
reinterpret_cast<TtsEngine* (*)()>(dlsym(engine_lib_handle, "getTtsEngine"));
@@ -362,18 +397,19 @@ android_tts_SynthProxy_native_setup(JNIEnv *env, jobject thiz,
Mutex::Autolock l(engineMutex);
pJniStorage->mNativeSynthInterface->init(ttsSynthDoneCB);
}
result = TTS_SUCCESS;
}
// we use a weak reference so the SynthProxy object can be garbage collected.
pJniStorage->tts_ref = env->NewGlobalRef(weak_this);
// save the JNI resources so we can use them (and free them) later
env->SetIntField(thiz, javaTTSFields.synthProxyFieldJniData,
(int)pJniStorage);
initializeEQ();
env->SetIntField(thiz, javaTTSFields.synthProxyFieldJniData, (int)pJniStorage);
env->ReleaseStringUTFChars(nativeSoLib, nativeSoLibNativeString);
return result;
}
@@ -842,9 +878,13 @@ static JNINativeMethod gMethods[] = {
(void*)android_tts_SynthProxy_shutdown
},
{ "native_setup",
"(Ljava/lang/Object;Ljava/lang/String;)V",
"(Ljava/lang/Object;Ljava/lang/String;)I",
(void*)android_tts_SynthProxy_native_setup
},
{ "native_setLowShelf",
"(ZFFFF)I",
(void*)android_tts_SynthProxy_setLowShelf
},
{ "native_finalize",
"(I)V",
(void*)android_tts_SynthProxy_native_finalize

20
packages/TtsService/src/android/tts/SynthProxy.java
View File

@@ -32,6 +32,15 @@ import java.lang.ref.WeakReference;
@SuppressWarnings("unused")
public class SynthProxy {
// Default parameters of a filter to be applied when using the Pico engine.
// Such a huge filter gain is justified by how much energy in the low frequencies is "wasted" at
// the output of the synthesis. The low shelving filter removes it, leaving room for
// amplification.
private final static float PICO_FILTER_GAIN = 5.5f; // linear gain
private final static float PICO_FILTER_LOWSHELF_ATTENUATION = -18.0f; // in dB
private final static float PICO_FILTER_TRANSITION_FREQ = 1100.0f; // in Hz
private final static float PICO_FILTER_SHELF_SLOPE = 1.0f; // Q
//
// External API
//
@@ -40,8 +49,11 @@ public class SynthProxy {
* Constructor; pass the location of the native TTS .so to use.
*/
public SynthProxy(String nativeSoLib) {
Log.v(TtsService.SERVICE_TAG, "TTS is loading " + nativeSoLib);
boolean applyFilter = nativeSoLib.toLowerCase().contains("pico");
Log.v(TtsService.SERVICE_TAG, "about to load "+ nativeSoLib + ", applyFilter="+applyFilter);
native_setup(new WeakReference<SynthProxy>(this), nativeSoLib);
native_setLowShelf(applyFilter, PICO_FILTER_GAIN, PICO_FILTER_LOWSHELF_ATTENUATION,
PICO_FILTER_TRANSITION_FREQ, PICO_FILTER_SHELF_SLOPE);
}
/**
@@ -161,8 +173,10 @@ public class SynthProxy {
*/
private int mJniData = 0;
private native final void native_setup(Object weak_this,
String nativeSoLib);
private native final int native_setup(Object weak_this, String nativeSoLib);
private native final int native_setLowShelf(boolean applyFilter, float filterGain,
float attenuationInDb, float freqInHz, float slope);
private native final void native_finalize(int jniData);