 |
Description  |
|
|
CROSS-REFERENCE TO RELATED APPLICATION
This application claims priority of European Patent Application No.
99308221.3, which was filed Oct. 18, 1999.
1. Field of the Invention
This invention relates to digital communications apparatus.
2. Description of the Related Art
A desirable feature in digital communication apparatus is the ability to
record speech. In, for example, a digital mobile phone, this ability would
enable the phone to act as a telephone answering machine, or to record a
voice memo, or to record a conversation taking place over the telephone.
An important consideration, particularly in a digital mobile phone, is the
amount of memory that has to be provided for this purpose. Previous
attempts at providing speech recording in a digital mobile phone have not
made efficient use of memory space, and thus required either an
unacceptably large memory, or utilised a smaller memory which was
insufficient for some purposes.
SUMMARY OF THE INVENTION
According to one aspect of this invention there is provided digital
communications apparatus including a comfort noise estimator for providing
silence frames containing information representative of background
acoustic noise, a comfort noise generator for providing comfort noise for
simulating background acoustic noise, and speech record/playback means
adapted, on record, to store speech frames substantially only during the
presence of speech, to store one or more silence frames at the end of the
presence of speech, and to store data representative of the duration of
the absence of speech, and adapted, on playback, to provide as output
speech signals derived from the stored speech frames and, in dependence
upon the stored one or more silence frames, comfort noise from the comfort
noise generator for a duration represented by the stored data.
The apparatus may be a digital mobile phone including a transmitter adapted
to be switched on only for transmission of frames containing useful
information, and wherein the comfort noise generator is adapted to provide
comfort noise in dependence upon silence frames when no speech frames are
received.
The apparatus may include a speech encoder for encoding speech into speech
frames and for providing input to the comfort noise estimator for
providing said silence frames, whereby said speech frames stored by the
speech record/playback means are those encoded by said speech encoder, and
a speech decoder for decoding stored speech frames and for providing in
the output of the speech decoder comfort noise generated by said comfort
noise generator in dependence upon stored silence frames.
The digital mobile phone may be a Global System for Mobile Communications
(GSM) phone.
Said stored data may comprise a count of speech frames occurring during the
absence of speech.
Said stored data may comprise a respective dummy frame of minimal length
for each speech frame occurring during the absence of speech.
According to another aspect of this invention there is provided a method of
recording/playing back speech in digital communications apparatus, the
method including, for recording, storing speech frames substantially only
during the presence of speech, storing one or more silence frames,
containing information representative of background acoustic noise, at the
end of the presence of speech, and storing data representative of the
duration of the absence of speech, and, for playback, providing as output
speech signals derived from the stored speech frames and, in dependence
upon the stored silence frame or frames, comfort noise for simulating
background acoustic noise for a duration represented by the stored data.
Said stored data may comprise a count of speech frames occurring during the
absence of speech.
Said stored data may comprise a respective dummy frame of minimal length
for each speech frame occurring during the absence of speech.
BRIEF DESCRIPTION OF THE INVENTION
The invention will now be described by way of example with reference to the
accompanying drawings, in which:
FIG. 1 is a schematic diagram of a GSM telephone embodying the invention;
and
FIGS. 2 and 3 are schematic diagrams showing in greater detail the Audio
Encode function and the Audio Decode function, respectively, shown in FIG.
1.
DETAILED DESCRIPTION
During a normal telephone conversation, the participants alternate so that,
on the average, each direction of transmission is occupied roughly 50% of
the time. In the GSM system so called "discontinuous transmission" (DTX)
is specified as a mode of operation whereby the transmitter is switched on
only for those frames which contain speech or other useful information.
This has the advantages that the average interference level "on air" is
reduced, leading to better spectrum efficiency, and, in the mobile phone,
battery life is prolonged, or a smaller battery may be used for a given
operational duration. The achievement of the discontinuous transmission
mode of operation requires a voice activity detector on the transmit side
to detect absence of speech, an evaluation of the background acoustic
noise on the transmit side in order to transmit characteristic parameters
of the background noise to the receive side, and generation on the receive
side of a similar noise, i.e., so called comfort noise, during periods
when the radio transmission is cut.
The transmission of comfort noise characteristics to the receive side is
achieved by means of a special frame, a so called silence descriptor
frame, or SID frame. This frame is transmitted at the end of each speech
burst and serves as an "end of speech" marker for the receive side. In
order to update the comfort noise characteristics at the receive side, SID
frames are also transmitted at regular intervals during speech pauses.
This also serves the purpose of improving the measurement of the radio
link quality by the radio sub-system.
In its application to a GSM phone, the present invention makes use of the
voice activity detector, the comfort noise generator and various other
components which are already present in the phone for the purpose of the
discontinuous transmission mode of operation described above.
Referring now to FIG. 1, speech signals from normally-provided microphone
101 in a GSM handset 100 are passed via a filter 102 to an analog to
digital converter 103, the digital output of which is provided to an Audio
Encode Function block 104. The output of block 104, whose functions will
be described in more detail with reference to FIG. 2, is provided as input
to a Transmit Chain block 105 where the usual channel encoding,
interleaving and modulation are effected. The output of block 105 is
provided as input to a Radio block 106, comprising the usual radio
frequency (RF) sections such as the power amplifier and antenna, for
translation to an RF signal for transmission. Received signals at the
output of the Radio block 106 are fed as input to a Receive Chain block
107 where the usual equalisation and channel decoding operations are
performed to produce speech data and various flags indicative of the
quality of the received signal (i.e. whether there are any errors in the
speech and SID frames, and whether the speech data is a speech frame or a
SID frame). The speech data and flags are fed as input to an Audio Decode
function block 108. The output of block 108, whose functions will be
described with reference to FIG. 3, is fed via a switch 109 as input to a
digital-to-analog converter 110 whose analog output is fed via a filter
111 to an electroacoustic transducer 112, such as an earpiece or
loudspeaker.
The input to Audio Encode function block 104 and the output from Audio
Decode function block 108 are also fed as inputs to a summer 113 and to a
switch 114. The output of summer 113 is also fed as input to switch 114.
The output of switch 114 is fed as input to Audio Encode function block
104' which may be, and preferably is, a second instance of
already-provided Audio Encode function block 104 which, as will be
understood by those skilled in the art, is normally embedded in software
in the digital signal processor, or DSP (not shown), in the GSM handset.
Referring now to FIG. 2, the input to Audio Encode function block 104' is
fed to a Speech Encoder 201 which encodes the speech signals into speech
frames. Signals from Speech Encoder 201 are fed to a Voice Activity
Detector 202 which produces a signal, a so-called VAD flag, in response to
the absence of speech. Signals from Speech Encoder 201 are also fed to a
Comfort Noise Estimator 203 which produces SID frames which contains
information representative of the background acoustic noise.
The VAD flag from Voice Activity Detector 202, speech frames from Speech
Encoder 201 and SID frames from Comfort Noise Estimator 203 are fed to a
Transmit DTX and Control block 204 which provides at its output speech
data and a speech, or SP, flag indicative of whether the speech data
output is a speech frame or a SID frame. Until Voice Activity Detector 202
detects the absence of speech, the speech data output comprises speech
frames and the SP flag is indicative thereof. At the end of a speech burst
it typically takes a time equal to N (N greater than 1) speech frames to
compute a new SID frame. When, therefore, Voice Activity Detector 202
detects the absence of speech, the output of block 204 continues to
comprise speech frames until the new SID frame becomes available whereupon
it appears at the output of block 204 and the SP flag changes to indicate
same. An exception to this arises when, at the end of a speech burst, less
than M (M>1) frames have elapsed since the last SID frame was computed.
In this situation the last SID frame is provided and the SP flag changes
to indicate same. In effect, if the last SID frame is not too old, it is
treated as being current and is therefore used, rather than waiting for a
new one to be computed. This feature avoids the wait which would otherwise
occur for a new SID frame to be computed in the case where a short
background noise spike is taken for speech, thus reducing activity on the
air.
Returning now to FIG. 1, the operation of Audio Encode function block 104
is similar to the operation just described for block 104'. As will be
appreciated by those skilled in the art, block 104 may operate at half
rate, full rate or enhanced full rate, depending on operational
circumstances or requirements, whereas block 104' will normally operate at
full rate. Block 104 serves to provide for transmission to a remote
receiver, via Transmit Chain block 105 and Radio block 106, speech frames
and SID frames and an SP flag indicative of whether the frames are speech
or SID.
For record operation, the speech data and SP flag output of block 104' is
provided to a Record and Playback Control block 115 which serves to decide
what to store in a memory 116 via a microcontroller 117. Record and
Playback Control block 115 causes to be stored in memory 116 all frames
which are indicated as speech frames by the SP flag. When the SP flag
indicates a SID frame, the first such SID frame is also stored in memory
116. Thereafter, there is no need to store any more frames until the SP
flag again indicates a speech frame. When such an indication of a speech
frame is received, block 115 causes to be stored in memory 116 data
representative of the duration of the intervening absence of speech.
Preferably such data is a count of speech frames not stored in memory 116,
i.e. a count of speech frames in which speech was absent. As an
alternative to storing such a count, the stored data may take other forms,
such as a dummy frame of minimal length for each unstored frame. As will
be appreciated, storing speech frames only when speech is present leads to
a very much more efficient use of memory space.
During long periods of absence of speech it may be desirable periodically
to store in memory 116 updated SID frames, in which case the data
representative of the duration of the absence of speech, such as the
above-noted frame count, would be a count from the first stored SID frame
to the second stored SID frame, a count from the second stored SID frame
to the third stored SID frame, and so on, ending with a final stored count
from the last stored SID frame to the onset of speech.
For playback operation, Record and Playback Control block 115 causes the
contents of memory 116 to be sent to Audio Decode block 108'. Thus when a
speech frame is retrieved from memory 116 it is sent to block 108' with
the flags set to indicate speech with no errors. (At the output of Receive
Chain block 107, information regarding the presence or absence of errors
in the received signals is relevant for decoding purposes in block 108,
but for the playback function such error information is irrelevant, and so
for block 108' only the speech/SID field is used with the flags set to
indicate the absence of errors). When a SID frame is retrieved from memory
116 it is sent to the Audio Decode block 108' with flags set to indicate a
SID frame with no errors. The same SID frame is repeatedly sent to block
108' for a period governed by the data stored in memory 116 representative
of the duration of the absence of speech, and this stored data is changed
to reflect the remaining duration of absence of speech. Thus, where such
data is a frame count, the count is reduced by 1 each time a SID frame is
sent to block 108'. Where, as described above, during a speech absence
more than one SID frame is stored in memory 116, the first SID frame is
sent until the first stored count reduces to zero, then the second stored
SID frame is sent until the second stored counts reduces to zero, and so
on.
Referring now to FIG. 3, Audio Decode function block 108' may be, and
preferably is, a second instance of already-provided Audio Decode function
block 108 (FIG. 1) which, as will be understood by those skilled in the
art, is normally embedded in software in the DSP (not shown) in the GSM
handset. Accordingly although the description of the operation is given
with reference to block 108', it will be understood that block 108 (FIG.
1) functions in a similar manner.
Speech data and flags at the input of block 108' are provided as input to a
Receive DTX and Control Function block 301 which passes speech frames,
indicated as such by the flags, directly to a speech decoder 302, the
output of which is the output of block 108'. One or more SID frames,
indicated as such by the flags, are sent to a comfort noise generator 303
which generates comfort noise whose characteristics depend upon
information contained in the SID frame or frames. The comfort noise is
provided to speech decoder 302 for the duration of the absence of speech
as represented by the stored data in memory 116 (FIG. 1) e.g. the count of
the frames during which speech was absent.
Thus the output of block 108' comprises the decoded speech frames that are
stored in memory 116 (FIG. 1) with each silence interval between speech
bursts containing comfort noise generated by comfort noise generator 313,
in dependence on the SID frame or frames stored in memory 116, for a
duration representated by the data, e.g. the frame count, stored in memory
116 representing the duration of the silence interval. Block 108' also has
a Speech Frame Substitution block 304 which is not used in the playback
function. It is merely a counterpart of a corresponding block in the first
instance of Audio Decode block 108 (FIG. 1) where it serves to insert one
or more substitute speech frames, or provide a "mute" output, if bad data
is received from the Receive Chain block 107.
Referring to FIG. 1 again, the output of block 108' is fed to
digital-to-analog converter 110 via switch 109. It will be appreciated
that, for normal operation as a telephone, switch 109 connects the output
of Audio Decode function block 108 to the input of digital-to-analog
converter 110, whereas, to listen to a recording, switch 109 connects the
output of Audio Decode function block 108' to the input of converter 110.
It will be appreciated that the setting of switch 114 controls what is
recorded. Thus by setting switch 114 to connect the output of
analog-to-digital converter 103 to the input of Audio Encode block 104',
the handset can record a voice memo, or record outgoing speech during a
telephone call, or record a voice prompt for prompting the user of the
handset, e.g. a prompt such as "Please speak the name of the person you
wish to dial now, or say "Help" for further options". With the switch 114
set to provide output from Audio Decode block 108 to the input of Audio
Encode block 104', the handset can record incoming speech during a
telephone call, or when acting as a telephone answering machine. When
switch 114 connects the output of summer 113 to the input of Audio Encode
block 104', recording of a telephone conversation, i.e. both incoming and
outgoing speech, is possible.
When the invention is applied to a radio transceiver employing
discontinuous transmission, various items of hardware or software that
already exist in the transceiver may be "re-used". Thus, in a GSM handset
all of the items shown in FIG. 1, with the exception of summer 113,
switches 109 and 114, and Record and Playback Control block 115, already
exist for the normal transmit and receive operations of the handset, and
so the invention is particularly efficient in its re-use of such hardware
and/or software.
As will be appreciated by those skilled in the art, the various items shown
in FIG. 1 may be implemented in hardware or software as appropriate.
Normally, microphone 101, filter 102, analog-to-digital converter 103,
Radio 106, the modulation finction in Transmit Chain 105,
digital-to-analog converter 110, filter 111, transducer 112 and memory 116
are implemented in hardware. Other functions in the Transmit Chain 105,
the Receive Chain 107, Audio Encode functions 104 and 104' and Audio
Decode functions 108 and 108' are normally implemented in software in the
DSP in the handset. The Record and Playback Control block 115 may be
implemented in software in the DSP, or in the microcontroller 117, or in
hardware.
Although the invention has been described by way of its application to a
GSM phone, it will be appreciated that it will find application in other
systems employing discontinuous transmission, such as PCS 1900 and IS-136
for example, and in systems where lower rate transmission occurs during
periods of silence, for example as may occur in CDMA systems. The
invention may also find application in so-called Voice over IP (Internet)
where a GSM EFR (Enhanced Full Rate) speech coder may be used to send
packet data over the internet, or a GPRS terminal able to act as a GSM
terminal as well as a data terminal may be used to transfer packet data
containing speech frames, a GSM speech encoder being used to encode the
speech.
* * * * *
|
|
 |
|
 |
Description |
|
