JP2017204840A - Voice quality estimation device, voice quality estimation method and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a voice quality estimation technique using a more precise sound loss rate, when estimating the hearing quality based on packet capture.SOLUTION: In a voice quality estimation device having a sound loss rate estimation unit for estimating sound loss rate from a sound packet flow, and a hearing quality estimation unit for estimating the hearing quality from the sound loss rate thus estimated, by using a predetermined mapping function, the sound loss rate estimation unit estimates the number of sound packet loss according to which of the combination of a tone packet or a silent packet, the packets before and after the lost packet in the sound packet flow belong, and estimates the sound loss rate from the number of sound packet loss thus estimated.SELECTED DRAWING: Figure 3

Description

  The present invention relates to a voice quality estimation apparatus, method, and program for a voice call service via a network.

  It is important for service providers to understand the quality of services in order to maintain and improve voice call services. One of the indicators for evaluating service quality is conversation quality. The conversation quality represents the quality experienced by the user using the service when having a conversation. In a scene of operation, since it is expensive and time consuming to evaluate the conversation quality by subjective evaluation, the conversation quality is estimated using the listening quality which is the main factor of the conversation quality. Here, the listening quality represents the quality that the user feels when listening to the voice.

  Listening MOS (Mean Opinion Score) is an evaluation index of listening quality. The evaluation of the listening MOS is actually judged by the evaluator listening to the audio sample, but there is an algorithm method that estimates without requiring the evaluator. For example, a technique based on POLQA (Perceptual Objective Listening Quality Analysis) is known as a technique for estimating the listening MOS of a broadband voice service such as VoLTE (Voice over Long Term Evolution) (see Non-Patent Document 1). POLQA is an objective evaluation method for evaluating listening quality by comparing a reference voice signal input from the utterance side with a recorded voice signal output on the reception side and calculating a POLQA evaluation value. The POLQA evaluation value is ITU-T (Telecommunication standardization sector of International Telecommunication Union) recommendation P.I. By applying a mapping function defined in 863 Implementer's guide (P. Imp 863), conversion to an estimated listening MOS (MOS-LQO: Mean Opinion Score-Listening Quality Objective) is possible.

  POLQA needs to input and process audio signals. However, the recording function is not always installed in the terminal, and it may be difficult to obtain a voice signal for the purpose of grasping the quality of the voice call service. In this way, voice signals cannot be handled directly, and there is a disadvantage that POLQA cannot be applied when estimating quality by capturing packets in the network.

  On the other hand, Non-Patent Documents 2 and 3 propose a method for estimating quality with high accuracy using packet capture. Non-Patent Document 2 and Non-Patent Document 3 use the characteristics of a codec (AMR-WB, etc.) that has a voice section detection mechanism, determine voice / silent sections according to the packet format, , Voice packet) is estimated in consideration of the loss rate (hereinafter referred to as voice loss rate).

  However, Non-Patent Document 2 is a full reference model that compares packet information after loss occurs in a network and the like and packet information before the loss occurs. There is an inconvenience that it cannot be applied.

  On the other hand, Non-Patent Document 3 is a non-reference model that is estimated from only packet information on the receiving side, and performs estimation in consideration of the sound loss rate based on packet capture at one point. In this estimation method, it is determined from the header information of the packets before and after the lost packet and whether or not the lost packet is a voice packet. In particular, when the preceding and following packets are voice packets and silence packets, all lost packets are regarded as voice packets. However, there are many situations where the voiced and silent sections are switched, because there are many situations where it is difficult to affect the listening MOS, or the listening MOS is actually high, but it may be underestimated. .

ITU-T P.863 Perceptual Objective Listening Quality Assessment., 09/2014. Okuyama Takafumi, Kurashima Atsuko, Masuda Yuki, "A Study on Estimation Method of Listening Quality in VoLTE," IEICE Societies, B-11-21, pp.438, March 2015 Yang F., Jiang L. and Li X. (2012) .Real-time quality assessment for voice over IP.Con-currency and Computation: Practice and Experience, 24 (11), 1192-1199.

  As described above, in a system that performs conversation by voice communication between a plurality of terminals via a network, when performing estimation considering the voice loss rate based on packet capture at one point, the voiced and silent sections are When packet loss occurs at the switching point, there is a problem that although the listening MOS is actually high, it may be underestimated.

  In order to solve the above-described problem, an object of the present invention is to provide a voice quality estimation technique using a more accurate voice loss rate when estimating listening quality based on packet capture.

  In order to solve the above-described problem, an aspect of the present invention provides a sound loss rate estimator that estimates a sound loss rate from a voice packet flow, and the estimated sound loss rate using a predetermined mapping function. A voice quality estimation device having a voice quality estimation unit for estimating a voice quality from the voice packet flow rate, and the voice loss rate estimation unit includes voice packets and silence packets before and after a lost packet in the voice packet flow. It is related with the audio | voice quality estimation apparatus which estimates the number of voice packet loss according to which combination and and estimates the said voice loss rate from the estimated number of voice packet loss.

  According to another aspect of the present invention, a voice packet flow is obtained, a voice loss rate is estimated from the voice packet flow, and a predetermined mapping function is used to calculate the voice loss rate from the estimated voice loss rate. A voice quality estimation method comprising: a step of estimating a listening quality, wherein the step of estimating the voice loss rate includes: a packet before and after a lost packet in the voice packet flow is a voice packet and a voice packet; The present invention relates to a method for estimating the number of voice packet losses according to which combination is used, and estimating the voice loss rate from the estimated number of voice packet losses.

  Still another aspect of the present invention relates to a program for causing a processor to function as each unit of the speech quality estimation apparatus described above.

  ADVANTAGE OF THE INVENTION According to this invention, when estimating listening quality based on a packet capture, the speech quality estimation technique using a more accurate voice loss rate can be provided.

FIG. 1 is a schematic diagram showing a configuration of a voice communication service via a network according to an embodiment of the present invention. FIG. 2 is a block diagram showing a configuration of a communication system according to an embodiment of the present invention. FIG. 3 is a block diagram showing a functional configuration of a speech quality estimation apparatus according to an embodiment of the present invention. FIG. 4 is a schematic diagram illustrating a procedure for estimating the number of voiced packets according to an embodiment of the present invention. FIG. 5 is a flowchart showing the voice quality estimation process according to one embodiment of the present invention.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  The voice quality estimation apparatus of the present invention receives a packet including voice communicated between terminals in a voice call system such as VoLTE composed of two voice call terminals connected via an IP network or a mobile network. It is measured at the side terminal or network.

  First, an outline of a voice communication service according to an embodiment of the present invention will be described with reference to FIG. FIG. 1 is a schematic diagram showing a configuration of a voice communication service via a network according to an embodiment of the present invention.

  As shown in FIG. 1, in the voice communication service according to the present embodiment, a wireless terminal performs voice communication via a carrier network (NW). More specifically, each wireless terminal wirelessly accesses a nearby base station in the carrier NW, and the base station exchanges a voice packet flow via the relay node. In this embodiment, the voice packet flow protocol uses RTP (Realtime Transport Protocol) including a sequence number and a transmission time stamp. As shown in the figure, the voice packet flow is composed of a silent packet and a voice packet. . According to RTP, silence packets have a relatively small size, and the spacing between silence packets is arranged to be relatively large, whereas voice packets have a relatively large size, and The intervals between voice packets are arranged to be relatively small. However, the voice communication service according to the present invention is not limited to RTP, and other protocols including a sequence number and a transmission time stamp may be used. For example, a codec (AMR-WB or the like) that has a voice section detection mechanism and has different packet formats and packet transmission intervals depending on sound / silence may be used.

  Next, a communication system according to an embodiment of the present invention will be described with reference to FIG. FIG. 2 is a block diagram showing a configuration of a communication system according to an embodiment of the present invention.

  As illustrated in FIG. 2, the communication system 10 includes a terminal 20, a network 30, a packet capture device 40, and a voice quality estimation device 100.

  The terminal 20 is a terminal having a packet-based voice communication function such as a VoLTE compatible terminal. The user of the terminal 20 can perform a conversation by voice communication via the network 30.

  The network 30 is a packet switching network such as a carrier network or the Internet. The voice packet flow transmitted from the terminal 20 is routed through a relay node in the network 30 and transmitted to the destination terminal 20.

  The packet capture device 40 acquires the voice packet flow transmitted from the transmission side terminal 20. In the illustrated embodiment, the packet capture device 40 acquires a voice packet flow between the receiving terminal 20 and the network 30. However, the present invention is not limited to this, and the packet capture device 40 may acquire the voice packet flow at any arbitrary point between the terminals 20. The packet capture device 40 transmits the acquired voice packet flow to the voice quality estimation apparatus 100. Further, the packet capture device 40 may be realized as a stand-alone device as shown in the figure, or may be mounted on the voice quality estimation device 100.

  The voice quality estimation apparatus 100 performs processing to be described later on the voice packet flow received from the packet capture apparatus 40, and outputs an estimated value of listening quality that is one index of voice quality.

  The voice quality estimation apparatus 100 may typically be realized by a server, and includes, for example, a drive device, an auxiliary storage device, a memory device, a processor, an interface device, and a communication circuit that are interconnected via a bus. . Various computer programs including programs for realizing various functions and processing described later in the voice quality estimation apparatus 100 are recorded on a recording medium such as a CD-ROM (Compact Disk-Read Only Memory), a DVD (Digital Versatile Disk), and a flash memory. May be provided. When the recording medium storing the program is set in the drive device, the program is installed from the recording medium to the auxiliary storage device via the drive device. However, it is not always necessary to install the program using a recording medium, and the program may be downloaded from any external device via a network or the like. The auxiliary storage device stores the installed program and also stores necessary files and data. The memory device reads and stores the program and data from the auxiliary storage device when there is an instruction to start the program. The processor executes various functions and processes of the speech quality estimation apparatus 100 described later according to various data such as a program stored in the memory device and parameters necessary for executing the program. The interface device is used as a communication interface for connecting to a network or an external device. The communication circuit executes various communication processes for communicating with a network such as the Internet. However, the hardware configuration described above is merely an example, and the speech quality estimation apparatus 100 is not limited to the hardware configuration described above, and may be realized by any other appropriate hardware configuration.

  Next, a speech quality estimation apparatus according to an embodiment of the present invention will be described with reference to FIG. FIG. 3 is a block diagram illustrating a speech quality estimation apparatus according to an embodiment of the present invention.

  As shown in FIG. 3, the speech quality estimation apparatus 100 includes a sound loss rate estimation unit 110 and a listening quality estimation unit 120.

  The voice loss rate estimation unit 110 estimates the voice loss rate from the voice packet flow. Here, the voice loss rate represents the ratio of the number of lost voice packets to the number of voice packets in the voice packet flow. That is, the sound loss rate L is

により算出可能である。ここで、ｐ_ｌｏｓｔは損失した有音パケットの個数であり、ｐ_ｒｃｖは受信した有音パケットの個数である。

Can be calculated. Here, p _lost is the number of _lost voice packets, and p _rcv is the number of received voice packets.

  In order to estimate the number of lost voice packets, the voice loss rate estimation unit 110 first detects a packet loss portion in the voice packet flow, and estimates the number of voice packets included in the detected packet loss portion. For example, the packet loss part may be detected based on the sequence number in the header information of each packet. That is, the sound loss rate estimation unit 110 may determine that a packet corresponding to the missing sequence number is a lost packet when a missing is detected in the sequence number of the received packet. Alternatively, the voice loss rate estimation unit 110 may determine the packet loss part from the progress status of the packet transmission time stamp in the voice packet flow. For example, when a difference between transmission time stamps larger than the larger one of the interval between voice packets and the interval between silence packets is detected for two packets received sequentially, the voice loss rate estimation unit 110 It may be determined that there is a packet loss portion between two packets.

  When the packet loss part is detected in this way, the voice loss rate estimation unit 110 detects whether there is a lost packet depending on which combination of the voice packet and the silence packet the packets before and after the detected packet loss part are. Estimate the number of sound packets. Note that the sound loss rate estimation unit 110 may determine whether each packet is a sound packet or a sound packet based on the packet size. That is, the packet size of the voice packet and the voiceless packet is defined according to the protocol type being used, and the voice loss rate estimation unit 110 determines the packet size by determining the packet size of each packet. It can be determined whether the packet is a voice packet or a silent packet.

  Specifically, Case 1) When the packet before and after the lost packet is a voice packet, the voice loss rate estimation unit 110 may estimate the lost packet as a voice packet. That is, as shown in FIG. 4, when the packets before and after the lost packet are voice packets, the voice loss rate estimation unit 110 estimates that all the lost packets are voice packets. Good. Then, the voice loss rate estimation unit 110 may calculate the number of lost packets based on the sequence numbers of the packets before and after the lost packet, and may determine the lost number as the number of lost voice packets. .

  Next, Case 2) When the packets before and after the lost packet are silence packets, the voice loss rate estimation unit 110 may estimate the lost packet as a silence packet. That is, as shown in FIG. 4, when the lost packet is a silent packet, the voice loss rate estimation unit 110 estimates all the lost packets as silent packets. In this case, the sound loss rate estimation unit 110 determines the number of lost sound packets as 0.

  Next, Case 3) When the packet before the lost packet is a voice packet and the subsequent packet is a voice packet, the voice loss rate estimation unit 110 estimates the lost packet as a voice packet. Also good. That is, as shown in FIG. 4, when the packet before the lost packet is a voice packet and the subsequent packet is a silence packet, the voice loss rate estimation unit 110 displays all the lost packets. It may be estimated that the packet is a silent packet. In this case, the sound loss rate estimation unit 110 determines the number of lost sound packets as 0.

  Next, Case 4) When the packet before the lost packet is a silent packet and the subsequent packet is a voice packet, the voice loss rate estimation unit 110 determines the sequence number and transmission time stamp of the lost packet. Based on the above, the number of lost voice packets may be estimated. Specifically, the sound loss rate estimation unit 110 calculates the number of lost packets based on the sequence numbers of the packets before and after the lost packet, and calculates the difference between the transmission time stamps of the packets before and after the lost packet. calculate. For example, in the specific example shown in FIG. 4, since the sequence numbers of the packets before and after the lost packet are “120” and “123”, respectively, it can be understood that there are two lost packets. Further, since the transmission time stamps are “0: 00: 00.0100” and “0: 00: 00.0400”, respectively, it can be seen that the difference between the transmission time stamps is 300 ms. At this time, the sound loss rate estimation unit 110

を満たす整数ｘを損失した有音パケットの個数として決定してもよい。ただし、Ｎは損失したパケットの個数であり、Ｔは送信タイムスタンプの差分であり、ｔ_０は無音パケット送信間隔であり、ｔ_１は有音パケット送信間隔であり、ｘ（≦Ｎ）は有音パケットの個数である。利用されるプロトコルタイプに応じて、無音パケット送信間隔及び有音パケット送信間隔は規定される。なお、ｘの値が複数存在する場合、有音損失率推定部１１０は、任意の１つを選択してもよい。

An integer x satisfying the above may be determined as the number of lost voice packets. Where N is the number of lost packets, T is the difference in transmission timestamps, t ₀ is the silent packet transmission interval, t ₁ is the voice packet transmission interval, and x (≦ N) is present. The number of sound packets. Depending on the protocol type used, the silence packet transmission interval and the voice packet transmission interval are defined. When there are a plurality of values of x, the sound loss rate estimation unit 110 may select any one.

  When the number of lost voice packets is determined in this way, the voice loss rate estimation unit 110 can determine the voice loss rate L according to the above-described equation.

  The listening quality estimation unit 120 estimates the listening quality from the estimated sound loss rate using a predetermined mapping function. The predetermined mapping function is any appropriate relational expression indicating the correspondence relationship between the sound loss rate and the listening quality estimation value. For example, the sound loss rate was given to various terminals by a prior experiment. It may be a relational expression indicating a correspondence relationship between the listening quality estimation value MOS_LQO measured by POLQA in a situation and a given sound loss rate.

  More specifically, the mapping function is obtained by a prior experiment, and in the quality degradation environment (e.g., the verification environment where the emulator is installed), the target terminal is used to perform in-network / terminal packet acquisition and listening quality measurement using POLQA. Further, it may be determined by curve fitting of a functional expression based on a combination (preferably with a plurality of variations) of the sound loss rate and the listening quality estimation value MOS_LQO (POLQA). For example, the mapping function is

であってもよい。ここで、MOS_LQOは受聴品質推定値であり、p1, p2はカーブフィッティングの精度が良い関数形状となる値であり、それぞれ、基準とする端末やコーデックにより実験的に定める値である。

It may be. Here, MOS _LQO is a listening quality estimation value, and p1 and p2 are values that have a function shape with good curve fitting accuracy, and are values that are experimentally determined by a reference terminal or codec, respectively.

  However, the mapping function according to the present invention is not limited to this, and may be any function that can be accurately fitted to the correspondence relationship between the sound loss rate and the listening quality estimation value obtained through experiments, for example, a quadratic function, The linear function may be a non-linear function.

  Next, a speech quality estimation process according to an embodiment of the present invention will be described with reference to FIG. FIG. 5 is a flowchart showing the voice quality estimation process according to one embodiment of the present invention.

  As shown in FIG. 5, in step S <b> 101, the sound loss rate estimation unit 110 acquires a voice packet flow. Specifically, the voice loss rate estimation unit 110 acquires a voice packet flow from the packet capture device 40.

  In step S102, the voice loss rate estimation unit 110 estimates the voice loss rate from the voice packet flow. Specifically, the voice loss rate estimation unit 110 estimates the number of voice packet losses depending on which combination of voice packets and voice packets the packets before and after the lost packet in the voice packet flow are. The voice loss rate is estimated from the estimated number of voice packet losses. As described above, the sound loss rate estimation unit 110 estimates that the lost packet is a sound packet when the packet before and after the case 1) lost packet is a sound packet, and the case 2) of the lost packet If the preceding and following packets are silence packets, the lost packet is assumed to be a silence packet. Case 3) If the packet before the lost packet is a voice packet and the subsequent packet is a silence packet, it is lost. Case 4) When the packet before the lost packet is a silent packet and the subsequent packet is a voice packet, the packet is estimated based on the sequence number of the lost packet and the transmission time stamp. The number of sound packets may be estimated. Here, in Case 4, the sound loss rate estimation unit 110

を満たす整数ｘを損失した有音パケットの個数として決定してもよい。ただし、Ｎは損失したパケットの個数であり、Ｔは送信タイムスタンプの差分であり、ｔ_０は無音パケット送信間隔であり、ｔ_１は有音パケット送信間隔であり、ｘ（≦Ｎ）は有音パケットの個数である。

An integer x satisfying the above may be determined as the number of lost voice packets. Where N is the number of lost packets, T is the difference in transmission timestamps, t ₀ is the silent packet transmission interval, t ₁ is the voice packet transmission interval, and x (≦ N) is present. The number of sound packets.

  In step S103, the listening quality estimation unit 120 estimates listening quality from the estimated sound loss rate using a predetermined mapping function. The mapping function may be any relational expression indicating the correspondence relationship between the sound loss rate and the listening quality estimation value, and may be defined by a prior experiment or the like.

  Note that each unit of the speech quality estimation apparatus 100 and steps S101 to S103 described above may be realized by a processor executing a program stored in a memory of a computer.

  As mentioned above, although the Example of this invention was explained in full detail, this invention is not limited to the specific embodiment mentioned above, In the range of the summary of this invention described in the claim, various deformation | transformation・ Change is possible.

DESCRIPTION OF SYMBOLS 10 Communication system 20 Terminal 30 Network 40 Packet capture apparatus 100 Voice quality estimation apparatus 110 Sound loss rate estimation part 120 Listen quality estimation part

Claims (7)

A voice loss rate estimator that estimates the voice loss rate from the voice packet flow;
A listening quality estimation unit that estimates listening quality from the estimated sound loss rate using a predetermined mapping function;
A speech quality estimation device comprising:
The voiced loss rate estimation unit estimates the number of voiced packet losses according to which combination of voiced packets and voiceless packets before and after the lost packet in the voice packet flow, and the estimation A speech quality estimation apparatus for estimating the speech loss rate from the number of speech packet loss. The sound loss rate estimation unit
i) When packets before and after the lost packet are voice packets, the lost packet is estimated as a voice packet;
ii) If the packets before and after the lost packet are silent packets, the lost packet is estimated as a silent packet;
iii) If the packet before the lost packet is a voice packet and the subsequent packet is a silence packet, the lost packet is estimated as a silence packet;
iv) When the packet before the lost packet is a silent packet and the subsequent packet is a voice packet, the number of voice packet loss is estimated based on the sequence number of the lost packet and a transmission time stamp. The speech quality estimation apparatus according to claim 1. The voice loss rate estimation unit iv) when the packet before the lost packet is a silent packet and the subsequent packet is a voice packet,

(Where N is the number of lost packets, T is the difference between the transmission time stamps of the preceding and succeeding packets, t ₀ is the silent packet transmission interval, t ₁ is the voice packet transmission interval, and x ( ≦ N) is the number of voice packet loss)
The speech quality estimation apparatus according to claim 2, wherein the voice packet loss number is estimated by obtaining an integer x satisfying   The speech quality estimation apparatus according to any one of claims 1 to 3, wherein the speech loss rate estimation unit determines whether each packet is a speech packet or a silence packet based on a packet size. The predetermined mapping function is:

(However, MOS _LQO = listening quality estimate, p1 and p2 are predetermined constants)
The speech quality estimation apparatus according to any one of claims 1 to 4, wherein Obtaining a voice packet flow;
Estimating a voiced loss rate from the voice packet flow;
Estimating listening quality from the estimated sound loss rate using a predetermined mapping function;
A speech quality estimation method comprising:
The step of estimating the voice loss rate estimates the number of voice packet loss according to which combination of the voice packet and the voice packet the packet before and after the lost packet in the voice packet flow is, A method of estimating the voice loss rate from the estimated number of voice packet losses.   The program for functioning a processor as each part of the audio | voice quality estimation apparatus as described in any one of Claims 1 thru | or 5.

Images