|
Hello everyone! this is regarding the usage of the "perceptual weighting filter" to improve the quality of synthesised speech.my query is referenced to the book "Digital Speech" by A.M.Kondoz the transfer function of the weighting filter, as given by Kondoz (page 145) is: W(z)= A(z)/A(z/gamma) ....where gamma lies between 0 &1 How does scaling by a const factor Gamma result in the desired frequency response for the filter (which emphasises the error signal in the spectral local minimas)? further.. (1)it is mentioned in the book that the effect of gamma is to 'broaden' the bandwidth of the formants by an amount given by: (-fs/pi)* ln(gamma) Hz (2)also, it is said that the center frequencies of the formants remain the same inspite of the scaling factor Gamma. Can anyone please explain the reasoning behind these two points? Thank you. Sameer Kibey Pune, India. |
|
|
Perceptual Weighting Filter
Started by ●January 12, 2002
Reply by ●January 12, 20022002-01-12
|
I can answer only the second querry. The formant ferquencies are nothing but the roots of A(z)=0; at z=exp(jw). the weighting filter is of the form W(z)=A(z)/A(z/gamma) . Its numerator has zeros same as formant frequencies and the roots of denominator are just scaled version of those of Nr. by (1/gamma). but at z=exp(jw) the, the w corresponding to these roots remains same as formant freqs. so only the magnitude gets affected, most at fomant frequencies. Thats why it is written in 'Kandoz' that "the amount of de-emphasis is controlled by gamma". -- On 12 Jan 2002 10:34:02 -000 speechcoding wrote: >_____________________________________ >Note: If you do a simple "reply" with your email client, only the author of this message will receive your answer. You need to do a "reply all" if you want your answer to be distributed to the entire group. > >_____________________________________ >About this discussion group: > >To Join: > >To Post: > >To Leave: > >Archives: http://www.yahoogroups.com/group/speechcoding > >Other DSP-Related Groups: http://www.dsprelated.com > >There is 1 message in this issue. > >Topics in this digest: > > 1. Perceptual Weighting Filter > From: Sameer Kibey < >________________________________________________________________________ >________________________________________________________________________ > >Message: 1 > Date: Sat, 12 Jan 2002 06:10:48 +0000 (GMT) > From: Sameer Kibey <> >Subject: Perceptual Weighting Filter > >Hello everyone! > >this is regarding the usage of the "perceptual >weighting filter" to improve the quality of >synthesised speech.my query is referenced to the book >"Digital Speech" by A.M.Kondoz > >the transfer function of the weighting filter, as >given by Kondoz (page 145) is: > >W(z)= A(z)/A(z/gamma) ....where gamma lies between 0 >&1 > >How does scaling by a const factor Gamma result in the >desired frequency response for the filter (which >emphasises the error signal in the spectral local >minimas)? further.. > >(1)it is mentioned in the book that the effect of >gamma is to 'broaden' the bandwidth of the formants by >an amount given by: > (-fs/pi)* ln(gamma) Hz > >(2)also, it is said that the center frequencies of the >formants remain the same inspite of the scaling factor >Gamma. > >Can anyone please explain the reasoning behind these >two points? > >Thank you. > >Sameer Kibey >Pune, India. > >________________________________________________________________________ >________________________________________________________________________ > >">http://docs.yahoo.com/info/terms/ > |
