6.3 High frequency band

26.1903GPPAdaptive Multi-Rate - Wideband (AMR-WB) speech codecSpeech codec speech processing functionsTranscoding functionsTS

For the higher frequency band (6.4 – 7.0 kHz), excitation is generated to model the highest frequencies. The high frequency content is generated by filling the upper part of the spectrum with a white noise properly scaled in the excitation domain, then converted to the speech domain by shaping it with a filter derived from the same LP synthesis filter used for synthesizing the down-sampled signal.

6.3.1 Generation of high-band excitation

The high-band excitation is obtained by first generating white noise uHB1(n). The power of the high-band excitation is set equal to the power of the lower band excitation u2(n) which means that

. ( 74)

Finally the high-band excitation is found by

, (75 )

where is a gain factor.

In the 23.85 kbit/s mode,is decoded from the received gain index.

In 6.60, 8.85, 12.65, 14.25, 15.85, 18.25, 19.85 and 23.05 kbit/s modes, gHB is estimated using voicing information bounded by [0.1,1.0]. First, tilt of synthesis etilt is found

(76 )

where is high-pass filtered lower band speech synthesis with cut-off frequency of 400 Hz. The is then found by

, ( 77 )

where gSP = 1 – etilt is gain for speech signal, gBG = 1.25gSP is gain for background noise signal, and wSP is a weighting function set to 1, when VAD is ON, and 0 when VAD is OFF. gHB is bounded between [0.1, 1.0]. In case of voiced segments where less energy is present at high frequencies, etilt approaches 1 resulting in a lower gain gHB. This reduces the energy of the generated noise in case of voiced segments.

6.3.2 LP filter for the high frequency band

6.3.2.1 6.60 kbit/s mode

The high-band LP synthesis filter AHB(z) is found by extrapolating the quantized ISF vector f into 20th order ISF vector fe. First, maximum of the autocorrelation Cmax(i) of ISF vector difference vector is obtained. Then new 16kHz ISF vector fe‘(i) is computed by

. (78 )

An approximation of the last element of new ISF vectoris updated based on lower frequency coefficients. New extrapolated ISF vector difference vector is

, (79 )

where cscale scales so that fe(19) will be equal to fe19. In order to insure stability, is bounded by

. (80)

Finally, the extrapolated ISF vector is obtained by

. (81 )

fe is converted to cosine domain to obtain with 16000 Hz sampling rate. The high-band LP synthesis filter AHB(z) is obtained by converting to LP filter as described in 5.2.4 with m=20.

6.3.2.2 8.85, 12.65, 14.25, 15.85, 18.25, 19.85, 23.05 or 23.85 kbit/s modes

The high-band LP synthesis filter AHB(z) is weighted low-band LP synthesis filter

, (82)

where Â(z) is the interpolated LP synthesis filter. has been computed analysing signal with the sampling rate of 12.8 kHz but it is now used for a 16 kHz signal. Effectively, this means that the frequency response FR16(f) of AHB(z) is obtained by

( 83)

where FR12.8(f) is the frequency response of A(z). This means that the band 5.1 – 5.6 kHz in 12.8 kHz domain will be mapped to 6.4 – 7.0 kHz in 16 kHz domain.

6.3.3 High band synthesis

uHB(n) is filtered through AHB(z). The output of this high-band synthesis sHB(n) is filtered through a band-pass FIR filter HHB(z) which has the passband from 6 to 7 kHz. Finally, sHB is added to synthesized speech to produce the synthesized output speech signal .