Speech digitizer

In a digital communication system operating in a multiframe format, a speech digitizer comprising:

• analyzer means connected to receive an analog speech waveform, said analyzer means including power and filter coefficient means responsive to said waveform for generating in digital format variable filter coefficient and power parameters representative of said waveform,
• pitch detector means responsive to said waveform for generating a digital pitch parameter substantially representing the fundamental periodicity of said waveform, said pitch detector means including;
• automatic gain control means for stabilizing said waveform,
• converter means for converting said analog waveform into a digital format in a predetermined time frame,
• function means for generating a digital signal representing an absolute magnitude difference function, said digital signal having a predetermined number of samples representing the variations in the pitch of said analog waveform and having a pattern of recurring maximum and minimum points over the pitch range,
• means for generating a first digital pitch signal representing the fundamental pitch of said waveform,
• periodicity means for generating a periodicity signal representing the ratio of one of said minimum and one of said maximum points,
• multiple check means connected to receive said digital signal and said first pitch signal for generating a second pitch signal when one of said multiple signals is lower than said first pitch signal by interpolating over successive ones of said samples to generate said second pitch signal, and
• range restrictor means connected to receive said digital signal, said periodicity signal, and said second pitch signal for restricting the range of said pitch signal to a range of pitches within a predetermined tolerance if the average pitch of said periodicity signal is below a predetermined level whereby a third pitch signal is generated representing the best estimate within the restricted range,
• voice detector means responsive to said waveform for generating a digital voicing parameter representing whether said speech waveform is voiced or unvoiced,
• multiplexer means for multiplexing said parameters into a digital serial data stream in said multiframe format where selected ones of the frames in said multiframe format occur as a synchronization frame,
• synchronization means for providing a digital synchronization code whereby said multiplexer multiplexes said synchronization code into a portion of said synchronization frame,
• first signaling interface means for connecting signaling information to another portion of said synchronization frame,
• means for transmitting said digital serial stream where said serial stream includes during said synchronization frame said synchronization code in one portion and said signaling information in said other portion,
• synthesizer means connected to receive said digital stream for generating a second analog waveform representative of said analog waveform, said synthesizer means including demultiplexer means for demultiplexing the transmitted parameters, the synchronization code, and the signaling information,
• second signaling interface means connected to receive the demultiplexed transmitted signaling information,
• periodic generator means for generating a periodic component signal representative of a pitch pulse signal and aperiodic generator means for generating an aperiodic signal representative of a random noise signal,
• mixer means connected to receive said component signals for mixing said component signals thereby forming a driving function signal, and
• filter means connected to receive said driving function signal for generating said second analog waveform thereby representing said first analog waveform.

In a digital communication system operating in a multiframe format, a pitch detector comprising:

• converter means connected to receive an analog speech waveform for converting said waveform to a digital format in a predetermined time frame corresponding to said multiframe format,
• means for generating a digital signal representing absolute magnitude difference function, said digital signal having a predetermined number of samples representing variations in the pitch of said waveform and having a pattern of recurring maximum and minimum points over the pitch period,
• means for generating a first digital pitch signal representing the fundamental pitch of said waveform,
• multiple check means connected to receive said digital signal and said first pitch signal for generating a second digital pitch signal when one of said recurring minimum points is lower than the first pitch signal by interpolating over successive ones of said samples of said digital signal to generate said second pitch signal thereby representing the fundamental pitch of said waveform,
• periodicity means for generating a periodicity signal representing the ratio of one of said minimum points and one of said maximum points, and
• range restrictor means connected to receive said digital signal, said periodicity signal, and said second pitch signal for generating a third pitch signal representing the restriction of the range of said first pitch signal to a range of pitches within a predetermined tolerance of the average pitch if said periodicity signal is below a predetermined level.

In a digital communication system operating in a multiframe format, a voicing detector connected to receive an analog speech waveform comprising:

• low pass detection and integration means for generating a low pass integrated signal representing the energy in a low frequency band of said waveform,
• high pass detection and integration means for generating a high pass integrated signal representing the energy in a high pass band of said waveform,
• first comparator means for generating a voicing function signal when the ratio of said low pass signal to said high pass signal exceeds a first predetermined threshold,
• second comparator means for generating a strong voicing function signal only when the ratio of said low pass signal to said high pass signal exceeds a second predetermined threshold and for generating a weak voicing signal when said ratio is less than or equal to said second threshold,
• third comparator means for comparing said low pass integrated signal with a filtered noise level signal of said low pass integrated signal representing background noise thereby forming a power present signal when said low pass integrated signal exceeds the noise level signal,
• means for determining the number of extreme maximum and minimum points occurring within a predetermined range in an absolute magnitude difference function level representing said waveform within a predetermined range thereby generating a structure number signal representing a voiced event when the number of extreme points is less than a predetermined number, and
• decision logic means for generating said voicing parameter in response to said strong voice signal, to said structure signal, to said periodicity signal, or to said weak voicing and said periodicity signal.

In a speech digitizer for use in a digital communication system operating in a multiframe format, the method comprising the steps of:

• generating in digital format in response to an analog speech waveform variable filter coefficient and power parameters representative of said waveform,
• generating a digital pitch parameter substantially representing the fundamental periodicity of said waveform,
• generating a digital voicing parameter representing whether said speech waveform is voiced or unvoiced,
• generating a digital signal representing an absolute magnitude difference function, said digital signal having a predetermined number of samples representing the variations in the pitch of said analog waveform and having a pattern of recurring maximum and minimum points over the frequency spectra,
• generating a first digital pitch signal representing the fundamental pitch of said sampled signal,
• generating a periodicity digital signal representing the ratio of one of said minimum and one of said maximum points,
• generating a second digital pitch signal when one of said recurring minimum points is lower than the first pitch signal by interpolating over successive ones of said digital signal to generate said second pitch signal,
• restricting the range of said pitch signal to a range of pitches within a predetermined tolerance of the average pitch if said periodicity signal is below a predetermined level whereby a third digital pitch signal is generated representing the best estimate with the restricted range,
• determining the number of extreme maximum and minimum points within a predetermined range of said difference function thereby generating a structure number signal representing a voiced event when the number of extreme points is less than a predetermined number,
• multiplexing said parameters into a digital serial data stream in said multiframe format where selected ones of the frames in said multiframe format occur as a synchronization frame,
• providing a synchronization code whereby said code is multiplexed into a portion of said synchronization frame,
• connecting signaling information to another portion of said synchronization frame,
• transmitting said digital serial stream,
• demultiplexing the transmitted parameters, the synchronization code, and the signaling information,
• receiving the demultiplexed transmitted signaling information,
• generating a periodic component signal representative of a random noise signal,
• mixing said component signals thereby forming a driving function signal, and
• generating a second analog signal thereby representing said first analog signal in response to said driving function signal.

In a pitch detector for use in a digital communication system operating in a multiframe format, the method comprising the steps of:

• converting an analog speech waveform to a digital format in a predetermined time frame corresponding to said multiframe format,
• generating a digital signal representing an absolute magnitude difference function of said waveform, said digital signal having a predetermined number of samples representing the variations in the pitch of said waveform and having a pattern of recurring maximum and minimum points over the pitch period,
• generating a first digital pitch signal representing the fundamental pitch of said waveform,
• generating a second pitch signal when one of said multiple signals is lower than the first pitch signal by interpolating over successive ones of said samples of said digital signal to generate said second pitch signal thereby representing the fundamental pitch of said waveform, and
• determining the number of extreme maximum and minimum points occurring within a predetermined range in an absolute magnitude different function signal representing said waveform thereby generating a structure number signal representing a voiced event when the number of extreme points is less than a predetermined number.

In a voicing detector connected to receive an analog speech waveform for use in a digital communication system operating in a multiframe format, the method comprising the steps of:

• generating a low pass integrated signal representing the energy in a low frequency band of said waveform,
• generating a high pass integrated signal representing the energy in a high pass band of said waveform,
• generating a voicing function signal when the ratio of said low pass signal to said high pass signal exceeds a first predetermined threshold,
• generating a strong voicing function signal only when the ratio of said low pass signal to said high pass signal exceeds a second predetermined threshold,
• generating a weak voicing signal if said ratio is less than or equal to said second threshold,
• comparing said integrated low signal with a filtered noise level signal representing background noise thereby forming a power present signal when the low pass integrated signal exceeds the noise level signal,
• determining the number of extreme maximum and minimum points occurring within a predetermined range in an absolute magnitude difference function signal representing said waveform thereby generating a structure number signal representing a voiced event when the number of extreme points is less than a predetermined number, and
• generating said voicing parameter in response to said strong voice signal, to said structure signal, to said periodicity signal, or to said weak voicing and said periodicity signal.