JPH06338908A - Sound packet assembly/disassembly device - Google Patents

Abstract

PURPOSE:To realize the constant sound quality of a large housing channel multiplicity by packet processing which summarizes plural channels. CONSTITUTION:Channel compression data 13 obtained by coding a PCM data string 11 by a sound coding/decoding part 2 by way of an interface part 1 is FIFO-operation-held by a sample waiting matrix part 3 and temporarily stored in correspondence to a sample by a sample collection part 4 to be bit data 15. Bit data 15 is blocked by each corresponding bit of each sample of a channel unit by a block processing part 5, added with packet header information 22 and generated to be a packet data string 17 by a line control part 6. On the other hand, the packet data string 17 is regenerated to be the CM data string 11 by the same reverse process. Packet processing bit kind information 20 is generated to make a packet processing bit deciding part 7 finish the packet processing of all the housed channels within a prescribed packet processing period unit in accordance with sound/silence identification information 19 of a period just before, and a channel managing table 8 registers and updates packet header information 22 by each packet processing period.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice PAD for converting voice media into a form capable of multimedia communication and multiplexing.
(Packet assembly / disassembly) device

[0002]

2. Description of the Related Art For example, reference (Saito et al .: Voice packet method, NTT Research Practical Report Vol. 38, No. 3, 231-2)
In the conventional voice PAD device shown in page 37, 1989), as shown in FIG. 15, the interface unit 1 connects the voice terminal such as a telephone via the PBX (Private Branch Exchange) with the incoming and outgoing voice PCM (pulse code modulation) data. Column 11 (see FIG. 16) A 30-channel time division multiplexed audio digital signal in which an analog audio signal is sampled at 8 kHz and subjected to 8-bit PCM. Each channel is identified by a time slot position in a 125 μs cycle frame, and TTC (telegraph The Telephone Technical Committee (based on the standard interface) and the channel PCM data 12 (64 kb / s voice PCM data for each channel) connected to the voice encoding / decoding unit 2 are delivered. The voice encoding / decoding unit 2 applies the voice signal to the voice signal in the voiced section for each channel.
The 8-bit channel PCM data 12 is encoded from the interface unit 1 by the adaptive differential PCM method or another method applied to a signal that cannot be discarded, and is generated as 5-bit channel compressed data 13 (40 kb / s compressed data for each channel). To do. Also, the channel compression data 13 is decoded from the packet disassembling / assembling unit 5a and reproduced as the original channel PCM data 12. The packet disassembling / assembling unit 5a
The channel compression data 13 is divided into packets (divided transmission units of information consisting of a header and a fixed-length block of the data information section) from the voice encoding / decoding unit 2 for each channel to generate channel packet data 16a. In addition, the channel packet data 16 from the packet multiplex synthesis / separation unit 5b.
The packet data information of a is assembled and reproduced as the original channel compressed data 13. Packet multiplexing / separation unit 5
In b, the channel packet data 16a is multiplexed and combined from the packet disassembling / assembling unit 5a to generate channel multiplexed packet data 16b. Also, the channel multiplexing packet data 16b is demultiplexed from the line control unit 6 and reproduced as each original channel packet data 16a. Line control unit 6
Is the channel multiplex packet data 16b connected to the packet multiplex / synthesis / separation unit 5b and the protocol (communication protocol) of the OSI (Open System Interconnection) reference model second (data link) layer. A transmission / reception voice packet data string 17 (CCITT) which is connected to a high-speed digital line connecting the multiplexer via the PMX of the own station.
(International Telegraph and Telephone Advisory Committee) Recommendation X. 21)).

The above-described conventional voice PAD device adopts a system (channel-unit parallel packet processing system) for disassembling packets for each channel and assembling them for packet processing.

As shown in FIG. 17, the packet disassembling / assembling unit 5a uses, for example, each channel 1 in time division multiplexing of 4 channels.
A packet processing interrupt is made every 4 to 4, and packet processing is performed by disassembling and assembling within a predetermined packet processing cycle unit (125 μs × N, for example, 16 ms when N = 128).

[0005]

The conventional voice PAD apparatus as described above adopts the channel unit parallel packet processing method in which each channel is disassembled and the assembled packet processing is performed. Therefore, the required channel scale is increased as the number of channels is increased. There was a problem to do.

The problem to be solved by the present invention is to aggregate a plurality of channels, block a predetermined bit of each channel unit sample, deblock it, and perform packet processing so as to increase the accommodation channel multiplicity and not deteriorate the voice quality. A method (packet processing method using a block corresponding to a predetermined bit of channel aggregation) is provided.

[0007]

A voice PAD device of the present invention is characterized by including the following means for solving the above-mentioned problems and adopting a packet processing system by a block corresponding to a channel aggregation predetermined bit.

The interface section delivers the input / output voice PCM data sequence and the channel PCM data connected to the voice encoding / decoding section.

The voice coding / decoding unit
Channel PCM data is encoded from the interface section and generated as channel compressed data. Also, the channel compression data is decoded from the sample queue section and reproduced as the original channel PCM data.

The sample queuing section is
Each sample of the channel compression data from the voice encoding / decoding unit and the sample data from the sample collecting unit is held by the first-in first-out operation.

The sample collecting unit temporarily stores, for each channel, each bit or each sample of the sample data from the sample queue unit and the bit data from the block processing unit in correspondence with the sample or with correspondence with the upper and lower bits.

The block processing unit refers to the packet processing bit type information from the channel management table for each channel, and blocks the bit data from the sample collection unit for each corresponding bit of each sample or each upper and lower bits of each channel unit channel management table. To generate a packet data string. Also, the line control unit deblocks the packet data information of the packet data string into corresponding bits or upper and lower bits of each sample of each channel and reproduces the original bit data.

The line control unit delivers the packet data sequence connected to the block processing unit and the incoming / outgoing voice packet data sequence.

According to the packet processing interrupt from the block processing unit, the packet processing bit determining unit discriminates voice / silence from the channel management table in the packet processing cycle immediately before the packet processing interrupt is performed for each channel on a packet processing cycle basis. Information, voice / silent identification information and signal type information, or voice / silence identification information, signal type information, and packet processing bit type information, is referred to within a predetermined packet processing cycle unit or a minimum packet processing cycle unit of a specific channel. In order to finish the packet processing of all the channels to be accommodated, all the bits are set for the relevant channel of the sound processing / soundless information in the immediately preceding packet processing cycle or the sound processing / soundless information is sound and the signal type information is the sound signal. If the upper bits or all of the bits and some of the upper bits and lower bits are used for packet processing, And generates a packet processing bit type information.

The channel management table registers and updates the packet processing bit type information from the packet processing bit determining unit and the control information from the block processing unit in correspondence with the channel number information as packet header information for each packet processing cycle.

[0016]

With the above-mentioned means, the voice PAD device of the present invention encodes the input outgoing voice PCM data string as channel PCM voice data (eg 64 kb / s) and compresses it as channel unit compressed data (eg 40 kb / s). To generate. Next, according to the packet processing bit type information for each channel, the compressed bit data, which is held for each channel by a FIFO (first in, first out) operation, and each bit is temporarily stored in correspondence with each sample or each sample in correspondence with the upper and lower bits, A packet is divided into blocks corresponding to corresponding bits or upper and lower bits of each unit sample, packet header information is added, and a generated voice packet data string is generated and output. The incoming voice packet data sequence to be input is reproduced and output as the incoming voice PCM data sequence in the same reverse process as above. On the other hand, in the packet processing bit determining unit, according to the voice / silent identification information, the voice / silent identification information and the signal type information or the voice / silent identification information, the signal type information, and the packet processing bit type information of the period immediately before the packet processing interrupt, In order to finish the packet processing of all channels accommodated within a predetermined packet processing cycle unit or within the minimum packet processing cycle unit of a specific channel, all bits and upper bits or The packet processing bit classification information is generated by classifying the packet processing bits according to some of the bits, upper bits, and lower bits. Further, in the channel management table, the packet processing bit type information and the control information such as the voiced / non-voiced identification information and the signal type information are registered and updated as packet header information for each packet processing cycle in correspondence with the channel number information.

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS As shown in FIG. 1, a voice PAD device according to an embodiment of the present invention has an interface unit 1, a voice encoding / decoding unit 2, and a line control unit 6 corresponding to FIG. The sample queuing section 3 FIFs each sample of the channel encoding data 13 from the voice encoding / decoding section 2 and the sample data 14 from the sample collecting section 4 for each channel.
It is held by the O (first in, first out) operation (see FIG. 2A). The sample collection unit 4 includes the sample queue unit 3 to the sample data 14 and the block processing unit 5 for each channel.
To bit data 15 are temporarily stored in correspondence with samples (see FIG. 2B). The block processing unit 5
The packet processing bit type information 20 is referred from the channel management table 8 for each channel, the bit data 15 is blocked from the sample collection unit 4 for each corresponding bit of each sample in each channel, and the packet header information 22 from the channel management table 8 is obtained. Add and packet data string 1
6 and connect to the line control unit 6. Further, the packet data information of the packet data string 16 is deblocked from the line controller 6 for each corresponding bit of each sample on a channel basis and reproduced as the original bit data 15 (FIG. 2).
(See (c)). In accordance with the packet processing interrupt 18 from the block processing unit 5, the packet processing bit determining unit 7 performs the packet processing interrupt for each channel in packet processing cycle units (see FIG. 3) in the immediately preceding packet processing cycle, and the channel management table 8 To voice / silence identification information 1
9, a predetermined packet processing cycle unit (125 μs ×
N, N is fixed), so that the packet processing time (T _TOTAL ) for all channels is calculated based on a predetermined criterion from the voice / silent channels so that the packet processing for all channels accommodated within The channel in which the voiced / non-voiced information of the cycle is voiced is classified into the case of packet processing of all bits and upper bits, and packet processing bit type information 20 is generated. In the channel management table 8, as packet header information 22 for each packet processing cycle, as shown in FIG. 4, the packet processing bit determining unit 7 and the packet processing bit type information 20 and the block processing unit 5 and the control information 21 (corresponding to the channel number information as shown in FIG. Voiced / unvoiced identification information and signal type information such as voice signal, push button signal, modem / FAX signal, etc.) are registered and updated.

The voice PAD apparatus of the above embodiment adopts a method (packet processing method by a channel aggregation predetermined bit corresponding block) of aggregating a plurality of channels and blocking and deblocking predetermined bits of each channel unit sample.

As shown in FIG. 5, the packet processing bit determining unit 7 first sets the channel pointer to the channel number 1 at the packet processing interrupt, clears the total packet processing time (T _TOTAL ) register of all channels to 0, and manages the channel. The voice / non-voice identification information 19 of the immediately preceding packet processing cycle is referred to from the table 8. If there is sound, all bits of the channel are packet processing bits and the total bit packet processing time (T _AB ) is added to T _TOTAL . If silence, and no packet processing bits in the channel, silent processing time (T _N) and T _AB and silence probability of continuity (P _N) virtual packet processing time in consideration of the _{(T VN = T N × P} N + T AB × (1-
P _N )) is added to T _TOTAL (above steps S1 to S
7). Steps S3 to S7 are repeated until all channels are completed (steps S8 and S9). Next, the calculated T _TOTAL is compared with a predetermined packet processing cycle unit (T _P ). If T _{TOTAL ≤T P} , it means that there is enough processing capacity, and packet processing is started. If T _TOTAL > T _P , the channel pointer is set to channel number 1 again, and the voice / silence identification information 19 is referred to. If voiced, the channel is (a M1 M2) Quality important upper bits when restoring the audio only a packet processing bits, T _AB upper bits packet processing time (T _HB) and the difference (T _AB the -T _HB) is subtracted from the T _TOTAL. Comparing the calculated T _TOTAL with T _P ,
_{If TOTAL} ≤ T _P , it means that there is a margin in processing capacity, and packet processing is started (above steps S10 to S1).
5). If T _TOTAL > T _P or there is no sound, steps S12 to S15 are repeated until the end of all channels (steps S16 and S1).
7). At the same time, a constraint condition is set in advance such as limiting the number of channels for packet processing. This is necessary in terms of CPU processing capacity. In the above, if there is voice on all channels from the voice / silence identification information 19 of the immediately preceding packet processing cycle, assuming that all bits are packet processing bits on all channels as shown in FIG. 6A, the total packet processing time T _{TOTAL of} all channels is predetermined. The packet processing cycle unit T _P is exceeded. As shown in FIG. 6B, if only the upper bits are used as the packet processing bits in all channels, the packet processing can be completed within a considerable margin within T _P. As shown in FIG. 6C, when only the upper bits in the channel numbers 1 to 3 and all the bits in the remaining channels are packet processing bits, the maximum channel multiplicity shown in FIG.
As compared with the case of (3), it is possible to realize the packet processing in which the quality at the time of restoring the voice is further improved. On the other hand, if the sound / silence identification information 19 of the immediately preceding packet processing cycle is plural channel silence,
Even if all the bits are used as packet processing bits in the voice channel as shown in FIG. 6D, the packet processing can be completed within T _P.
If the number of channels changing from silence to sound in the voice / silence identification information 19 of the immediately preceding packet processing period continues more than the previously predicted sound continuation probability (1- _PN ), the silence continuation probability P _N Perform processing such as changing settings.

In the above-mentioned embodiment, the packet processing bit determining unit 7 has a predetermined packet processing cycle unit (125 μs × N,
(N is fixed), so that the packet processing of all channels accommodated in () is terminated by classifying the channel having voiced / unvoiced information in the immediately preceding packet processing period as voiced by packet processing of all bits and upper bits. As described above, according to the packet processing interrupt 18 from the block processing unit 5, in the packet processing cycle immediately before the packet processing interrupt (see FIG. 3) in units of the packet processing cycle is performed for each channel, from the channel management table 8 / Refer to not only the silence identification information 19 but also the signal type information. As shown in FIG. 7, the voice / soundless information 19 of the immediately preceding packet processing cycle is voiced and the signal type information is a voice signal of that channel (voice / soundless). (Except when the information 19 is silent and when the signal type information cannot be discarded, such as push button signals and modem / FAX signals), all bits and upper bits It may be classified in the case of by the packet processing. In the above, the voice / silent identification information 19 and the signal type information of the immediately preceding packet processing cycle are the voice numbers for the channel numbers 1 to 3 and the push button signal or the modem / FAX signal, and the voice numbers and the voice signals for the channel numbers 4 to 7. , If the channel number 8 is silent, if packet processing is performed in this state as shown in FIG. 6E, the total packet processing time T of all channels is
_TOTAL exceeds the packet processing time T _P. Figure 6F
As described above, if only the upper bits are used as the packet processing bits in the channel numbers 4 and 5, the packet processing of all channels can be completed within T _P. As shown in FIG. 7, at the time of a packet processing interrupt, the signal type information of the immediately preceding packet processing cycle except for the relevant channel of the push button signal or the modem / FAX signal (step S12a), first, the total packet processing time T
_The respective means S1 to S9 for calculating _TOTAL and the respective steps S10 to S17 for classifying the voiced channel into all the bits and the packet processing bits of the upper bits correspond to FIG. 5 of the above embodiment.

Further, in the above embodiment, the packet processing bit determining unit 7 uses the packet processing interrupt 1 from the block processing unit 5.
In accordance with No. 8, the packet processing cycle unit for each channel (see FIG.
In the packet processing cycle immediately before the packet processing interrupt (see), the minimum packet processing of the specific channel is performed by referring to the packet processing bit type information as well as the voice / silent identification information 19 and the signal type information from the channel management table 8. In order to finish the packet processing of all channels accommodated in the cycle unit (125 μs × N, N is variable), the packet processing generation channel is analogized from the packet processing bit type information of the immediately preceding minimum packet processing cycle as shown in FIG. The channel with voice / silence information 19 of the immediately preceding packet processing cycle having voice and the signal type information having a voice signal may be classified by packet processing of all bits and upper bits. If the immediately before the packet processing period of voiced / silent identification information 19 is voiced and the signal type information is the audio signal, when the left packet processing in this state as shown in FIG. 10a, T _TOTAL
Since T _p exceeds T _P , packet processing of all channels can be completed within the minimum T _P if only the upper bits in the channel numbers 1 to 7 are used as packet processing bits as shown in FIG. In the case of FIGS. 10B and 10C, if only the upper bits in the channel number 1 are used as packet processing bits, the packet can be finished within the minimum T _P as shown in FIGS. 10B ′ and 10C ′. In the case of FIG. 10d, since only channel numbers 1, 4, and 7 are the packet processing generation channels, they can be finished within the minimum T _P. As shown in FIG. 9, at the time of packet processing interruption, each means S1 to calculate the total packet processing time T _TOTAL by analogizing the packet processing generation channel from the packet processing bit information of the immediately preceding packet processing cycle (steps S3a and S12b). S9 and the following steps S10 to S17 for classifying the voiced channel of the audio signal into all bits and the packet processing bits of the upper bits correspond to FIG. 7 of the above embodiment.

Further, in the above embodiment, the packet processing bit determining unit 7 uses the packet processing interrupt 1 from the block processing unit 5.
In accordance with No. 8, the packet processing cycle unit for each channel (see FIG.
And FIG. 8), in the immediately preceding packet processing cycle when the packet processing interrupt is performed, all bits are processed in the respective steps of FIG. 5, FIG. 7, and FIG. 9 of the above embodiment, as shown in FIG. 11, FIG. 12, and FIG. Instead of classifying according to the packet processing of the upper bits and the upper bits, a part of the upper bits and a part of the lower bits (M3 or M3 and M4) may be classified. 11 and 12 and 13,
At the time of packet processing interruption, a block Mi (i =
5) is set (step S11a). Next, a block Mi (i = i-1) that collects Mi bits is set as a block that does not perform packet processing, and the procedure S11 is performed as long as it is a block that collects lower bits (M3 or M3 and M4).
~ S16 is repeated (steps S16a and S16b).

Further, in the above embodiment, the sample collection unit 4 and the block processing unit 5 have been described as temporarily storing each bit corresponding to each sample for each channel, and blocking and deblocking each corresponding bit of each sample in each channel. However, as shown in FIGS. 14 (b) and 14 (c), each sample is sampled for each channel as the upper bit (M1 and M2) and the lower bit (M3).
Up to M5), temporary storage may be performed, and upper and lower bits of each channel unit sample may be blocked and deblocked.

[0024]

As described above, the voice PAD device of the present invention adopts a system in which a plurality of channels are aggregated and predetermined bits of each sample in each channel are blocked and deblocked to perform packet processing. There is no need to increase the required channel scale with the increase in the number of channels compared to the channel-by-channel parallel packet processing method that disassembles into packets and processes assembled packets, and only the high-order bits that are important when restoring voice to the voice during load concentration are processed. Therefore, there is an effect that a large accommodation channel multiplicity and a constant voice quality can be realized.

[Brief description of drawings]

FIG. 1 is a diagram showing an embodiment of voice packet assembly and
The functional block diagram of a decomposition | disassembly apparatus.

FIG. 2 is a diagram illustrating the functions of a sample queuing unit, a sample collecting unit, and a block processing unit shown in FIG.

FIG. 3 is a diagram for explaining a packet processing interrupt in units of fixed packet processing cycles for each channel of the packet processing bit determination unit shown in FIG.

FIG. 4 is a view for explaining registration / update packet header information of the channel management table shown in FIG.

5 is a flowchart illustrating the operation of a packet processing bit determining unit shown in FIG.

FIG. 6 is a diagram for explaining the effect of the operation of the packet processing bit determining unit shown in FIGS. 5 and 7.

FIG. 7 is a flowchart illustrating the operation of another embodiment of the packet processing bit determining unit shown in FIG.

8 is a diagram illustrating a packet processing interrupt in units of variable packet processing cycles for each channel of the packet processing bit determination unit shown in FIG.

9 is a flowchart illustrating the operation of another embodiment of the packet processing bit determination unit shown in FIG.

FIG. 10 is a diagram for explaining the effect of the operation of the packet processing bit determining unit shown in FIG.

FIG. 11 is a flowchart illustrating the operation of another embodiment of the packet processing bit determining unit shown in FIG.

12 is a flowchart illustrating the operation of another embodiment of the packet processing bit determining unit shown in FIG.

FIG. 13 is a flowchart illustrating the operation of another embodiment of the packet processing bit determining unit shown in FIG.

14 is a diagram for explaining the functions of another embodiment of the sample collection unit and the block processing unit shown in FIG.

FIG. 15 is a functional block diagram of a conventional voice packet assembling / disassembling apparatus.

FIG. 16 is a diagram for explaining the relationship between a 125 μs cycle frame 30-channel outgoing / incoming voice PCM data string, 64 kb / s channel PCM data, and 40 kb / s channel compressed data.

FIG. 17 is a diagram for explaining a packet processing interrupt in units of fixed packet processing cycles for each channel of the packet disassembling / assembling unit shown in FIG.

[Explanation of symbols]

1 interface section 2 voice encoding / decoding section 3 sample queue section 4 sample collection section 5 block processing section 6 line control section 7 packet processing bit determination section 8 channel management table 11 outgoing / incoming voice PCM data string (TTC standard interface) 12 Channel PCM data (64 kb / s) 13 Channel compressed data (40 kb / s) 14 Sample data 15 bit data 16 Packet data string 17 Originating and terminating voice packet data string (CCITT recommendation X.21 interface) 18 Packet processing interrupt 19 Voice / Silence identification information 20 Packet processing bit type information 21 Control information 22 Packet header information In the drawings, the same reference numerals indicate the same or corresponding parts.

Claims (5)

[Claims] 1. An interface unit for delivering input / output voice PCM data strings and channel PCM data connected to a voice encoding / decoding unit, and a channel for encoding channel PCM data from the interface unit for each channel. Generates compressed data, decodes the compressed channel data from the sample queue, and restores the original channel PC
The voice encoding / decoding unit for reproducing as M data, and the sample queue unit for holding each sample of channel compressed data from the voice encoding / decoding unit and sample data from the sample collecting unit for each channel in a first-in first-out operation. And a sample collection unit that temporarily stores each bit of sample data from the sample queue unit and bit data from the block processing unit for each channel in correspondence with samples, and packet processing bit type information from the channel management table for each channel. Referring to the bit data from the sample collection unit, the data is blocked for each corresponding bit of each channel unit sample, packet header information from the channel management table is added to generate a packet data sequence, and the line control unit generates a packet data sequence. Channel packet data information The block processing unit that deblocks each corresponding bit of each sample and reproduces it as the original bit data, and the packet data sequence connected to the block processing unit and the incoming and outgoing voice packet data sequence input and output are transferred. According to the packet processing interrupt from the line control unit and the block processing unit, in the packet processing cycle immediately before the packet processing interrupt for each channel is performed, the voice / silent identification information is output from the channel management table. In order to finish the packet processing of all channels accommodated within a predetermined packet processing cycle unit, the channel in which the voiced / silent information in the preceding packet processing cycle is voiced is processed by packet processing of all bits and upper bits. Packet processing bits that are classified into cases and generate the packet processing bit type information A channel management unit for registering and updating packet processing bit type information from the packet processing bit determination unit and control information from the block processing unit in correspondence with channel number information as the packet header information for each packet processing cycle. A voice packet assembling / disassembling device including a table. 2. The packet processing bit determining unit refers to not only the voice / silent identification information but also the signal type information from the channel management table to terminate the packet processing of all channels accommodated within a predetermined packet processing cycle unit. 3. The method according to claim 1, wherein the voice / silent identification information in the immediately preceding packet processing cycle is voice and the signal type information is a voice signal, and the channel is classified by packet processing of all bits and upper bits. Voice packet assembly / disassembly device. 3. A packet processing bit determining unit refers to not only voice / silence identification information and signal type information but also packet processing bit type information from a channel management table, and stores all packets within a minimum packet processing cycle unit of a specific channel. In order to end the packet processing of the channel, the packet processing generation channel is analogized from the packet processing bit type information of the immediately preceding minimum packet processing period, and the voice / silence information of the immediately previous packet processing period is voiced and the signal type information is voice. 3. The voice packet assembling / disassembling apparatus according to claim 2, wherein the channel of the signal is classified into packet processing of all bits and upper bits. 4. The packet processing bit determination unit classifies the packet processing bits according to all the bits and the upper bits and a part of the lower bits, instead of classifying the packets according to the packet processing of all the bits and the upper bits. The voice packet assembling / disassembling apparatus according to claim 1, 5. A sample collection unit waits for samples for each channel, and each sample of sample data from the matrix unit and bit data from the block processing unit is temporarily stored in correspondence with the upper and lower bits, and the block processing unit collects the samples. 5. The block data and the packet data information of the packet data string from the line control unit are blocked and deblocked for each upper and lower bits of each sample in each channel, and deblocked. Voice packet assembling / disassembling device.