The Field of the Invention

The invention relates to a speech decoding method, comprising the steps of receiving information frames containing speech encoding parameters, the first bits in each frame forming a synchronization word, and a synchronization check bit being thereafter repeated at regular intervals within the frame, synchronizing a speech decoder with the information frames received and extracting said speech encoding parameters from said information frames, storing the speech encoding para¬ meters until the following information frame is received, decoding an encoded speech signal by means of said stored speech encoding parameters, monitoring said synchronization check bits to detect errors occurring within an information frame.

The Background of the Invention The GSM (Global System for Mobile Communica¬ tions) is a pan-European mobile phone system, which is becoming a world-wide standard. Figure 1 shows the basic structural elements of the GSM system in a simplified manner without detailed description of their properties or the remaining parts of the system. A mobile exchange MSC switches incoming and outgoing calls. Most of its functions are similar _: to those of the exchange of a public switched telephone network. In addition to these normal functions, the mobile exchange also has functions typical of mobile communications, such as subscriber location management. Mobile stations MS are connected to the exchange MSC by means of base station systems. A base station system comprises a base station control¬ ler BSC and base stations BTS. Each of the base station controllers is used for controlling a plurality of base

stations. A BSC performs, for example, handovers in cases where the handover takes place between two base stations controlled by the same BSC.

The GSM system is entirely digital; even speech transmission is carried out entirely digitally. On account of the capacity restrictions of the radio path used, it is necessary to have an encoding method which is different from normal digital encoding methods and which takes into account the special features of speech; such a method is the RPE- TP method (Regular Pulse Excitation - Long Term Prediction), which utilizes both long and short term prediction. The encoding produces LAR, RPE and LTP parameters for transmission.

In the network, various speech encoding and rate adaption functions have been concentrated in one transcoder unit TRCU (Transcoder/Rate Adaptor Unit). The TRCU may have several alternative locations in the sys¬ tem, depending on the choices made by the manufacturer. When the transcoder TRCU is positioned apart from the base station BTS, information is transmitted between the BTS and the TRCU in so-called TRAU frames of 320 bits, meaning that at a transmission rate of 16 kbit/s it takes 20 ms to transmit one frame. There are four different types of frames, classified by their informa- tion content: speech, use/maintenance and data frames, and so-called idle speech frames. To enable synchroniza¬ tion, the first two octets in each frame contain 16 syn¬ chronization bits. In addition, the first bit of the 16-bit words (two octets) that form a frame is a syn- chronization check bit. In addition to data bits con¬ taining actual speech, data or use/maintenance informa¬ tion, all frames comprise control bits C1-C21, which transmit information on the type of the frame and a varying amount of other frame type specific information. In both types of the speech frames the last four bits

T1-T4 are reserved for timing alignment. Figure 2 illus¬ trates a speech frame according to GSM Recommendation 08.60. The data to be transmitted in the frame consists of speech encoding parameters. First there is a 36-bit group of LAR parameters. Then there are four subframes I, II, III and IV, each of which contains a 56-bit group of LTP-RPE parameters.

The interfaces of the transcoder are a 64 kbit/s PCM (Pulse Code Modulation) interface towards the mobile exchange MSC and a 16 kbit/s GSM interface towards the base station BTS. In connection with these interfaces the GSM recommendations also use the terms uplink and downlink direction: the uplink direction is the direction from the base station unit BTS towards the mobile exchange MSC, and the downlink direction is the reverse direction.

At the beginning of a call a transcoder TRCU is reserved for the call. Synchronization with the TRAU frames from the uplink direction is effected by means of the above-mentioned synchronization bits and syn¬ chronization check bits in such a manner that the actual synchronization with the frame takes place by means of the first 16 synchronization bits. The frame synchron¬ ization is effected continuously, and the time window used for monitoring is shifted according to possible changes in timing. The subsequent synchronization check bits in the frame are only used for detection of errors occurring within the frame. An error detected in the synchronization check bits does not call for action until in at least three consecutive erroneous TRAU frames are detected, in which case the level of the decoded speech is attenuated until synchronization is restored.

A handover between two base stations BTS can be performed either as an intra-BSC operation within a

base station controller BSC using the same PCM line to the exchange or as an inter-BSC operation between base station controllers by changing the PCM line to the exchange. In the existing system, the base station con- troller BSC performs handovers within a base station controller by switching a new base station in place of the old base station at a random moment in relation to the TRAU frames transmitted by the new base station. This is illustrated in Figure 3. The column on the left illustrates the TRAU frames transmitted by the old base station BTS1, the column in the middle illustrates the TRAU frames transmitted by the new base station BTS 2, and the column on the right illustrates the TRAU frames received by the transcoder. Base stations BTS1 and BTS2 are not synchronized with each other; the TRAU frames transmitted by them may thus have an arbitrary phase difference Δt. The transcoder TRCU is synchronized with the frame transmitted by the base station BTS1 at a point of time t _{l r} when receiving 16 synchronization bits "0" at the beginning of the frame. Thereafter it mon¬ itors synchronization bits "1", which regularly occur in the frame, and detects a proper synchronization op¬ eration. At a point of time t ₂ the base station control¬ ler BSC performs a change of base station and switches the transcoder TRCU to a new base station BTS2 while the transcoder is receiving a TRAU frame, whereby the last frame transmitted by the old base station is inter¬ rupted. The frame which is transmitted by the new base station BTS2 and whose reception begins at a random point within the frame is attached to the original interrupted frame in the transcoder. The data contained at the end of the last frame whose transmission was started by the old base station BTS1 is thus useless. In the existing systems the transcoder TRCU does not, however, take this into account but decodes the "speech

encoding parameters" included in this data. These para¬ meters may have any values, wherefore an undue disturb¬ ing sound effect is produced amidst normal speech due to the handover. An error is usually detected in the synchronization check bits, but it does not call for action, as in the following frame, transmitted by the new base station BTS2, the synchronization will be restored.

Disclosure of the Invention

An object of the present invention is to avoid the above-mentioned undue disturbing sound effect on a speech channel.

This is achieved with a speech decoding method as described in the introductory part of the specifica¬ tion, characterized in that, in response to detection of an error within an information frame, updating of the stored speech encoding parameters is prevented until the reception of following information frame. According to the invention, an error in the synchronization check bits is immediately taken into account in speech decoding by preventing the use of new useless speech encoding parameters for speech decoding and by using instead the stored speech encoding para- meters of the previous frame. In addition, the parameter controlling the level of speech can be modified in such a manner that the level of decoded speech is attenuated to a zero level or by a predetermined attenuation co¬ efficient. The invention also relates to a speech decoder, comprising means for receiving information frames con¬ taining speech encoding parameters, the first bits in each frame forming a synchronization word, and a syn¬ chronization check bit being thereafter repeated at regular intervals within the frame, means for synchron-

izing the speech decoder with the information frames received and for extracting said speech encoding para¬ meters from said information frames, means for storing said speech encoding parameters until the reception of the following information frame, means for decoding an encoded speech signal by means of the stored speech encoding parameters, and means for monitoring said syn¬ chronization check bits to detect errors occurring within the information frame. The invention is charac- terized by said monitoring means being responsive to the detection of a synchronization error within the informa¬ tion frame for preventing the updating of speech encoding parameters stored in storing means until the following information frame, starting from the speech encoding parameters during the reception of which the synchronization error was detected.

The invention also relates to the use of a speech decoder according to claims 8 to 11 in a cellular radio system for receiving information frames which are transmitted by a base station, said parameters being speech encoding parameters sent by a mobile station to the base station over a radio path, and said errors occurring within the frames being caused by a handover performed between base stations during a call.

A Brief Description of the Drawings

In the following the invention will be described in more detail with reference to the accom¬ panying drawings, in which Figure 1 illustrates the basic elements of the

GSM radio system,

Figure 2 illustrates a TRAU speech frame according to GSM Recommendation 8.60,

Figure 3 illustrates the effect of the handover on the transmission of TRAU frames, and

Figure 4 is a general block diagram of the speech decoder according to the invention.

Preferred Embodiments of the Invention The basic structural elements of the GSM system as well as the TRAU speech frame are described in GSM recommendations. As for a more accurate description of the GSM system, reference is made to GSM recommendations and "The GSM System for Mobile Communications" by M. Mouly and M-B. Pautet, Palaiseau, France, 1992, ISBN: 2-9507190-0-7.

Figure 4 is a general block diagram of the speech decoder according to the invention. The speech decoder performs all the necessary operations pertaining to speech decoding and to the processing of TRAU frames. Transmission of speech is dealt with under chapter 06 in the GSM Recommendations, speech encoding particularly under 06.10. The following will only deal with the improvement which the invention provides in the opera- tion according to the recommendations.

In Figure 4 a synchronization unit 41 identi¬ fies the beginning of a TRAU speech frame received from a base station BTS on the basis of the 16 synchron¬ ization bits in state "0" at the beginning of the frame, and is synchronized with the frame by means of these synchronization bits. Thereafter the unit 41 receives control bits C1-C15, a 36-bit group of LAR parameters, and four subframes I, II, III and IV, each of which contains a 56-bit group of LTP and RPE parameters. During reception, the synchronization unit 41 checks the 16-bit synchronization word at the beginning of the frame and thereafter that the first synchron¬ ization check bit of each received 16-bit word, i.e. the bit immediately following the first 16 synchronization bits and subsequently every sixteenth bit, is in logical

state "1". If all the checked synchronization check bits are "ones", a proper synchronization and an error-free frame are assumed by the unit 41. In this case the frame error output 48 is in an inactive state. When the unit 41 detects that a synchronization check bit is in state "0", it interprets this as an error within the frame. As a result the frame error output 48 will be set to an active state for the rest of the frame. When the unit 41 receives 16 synchronization zeros again at the beginning of the following frame, the frame error output 48 is reset to an inactive state. Such an error within a frame usually occurs due to an intra-BSC handover, as illustrated in connection with Figure 3.

The unit 41 extracts LAR parameters from the frame and feeds them to a buffer memory 43 of LAR para¬ meters. Correspondingly, the synchronization unit 41 extracts subframes I, II, III and IV, i.e. 56-bit groups of LTP and RPE parameters, from the received TRAU frame and supplies them to the buffer memories 44, 45, 46 and 47 of LTP-RPE parameters. Updating of parameters in the buffer memories 43, 44, 45, 46 and 47 is controlled by the frame error output 48 of the synchronization unit 41, the output being connected to the buffer memories. When the frame error output 48 is in an inactive state, the LAR parameters and the LTP-RPE speech encoding para¬ meters of the subframes I, II, III and IV are loaded to the buffer memories 43-47 as soon as they are received. From the buffer memories 43-47 the parameters are supplied to an LTP-RPE speech coder 42, which with the aid of the parameters decodes the encoded speech signal. Thus, upon receiving error-free TRAU frames, the para¬ meters are stored in the buffer memories 43-47 until they are updated by the parameters of a new TRAU frame. If an error is detected within a received TRAU frame, the frame error output 48 is immediately set to an

active state, as stated above. The active frame error output 48 thus prevents the buffer memories 43-47 from being updated by the parameters of an erroneous TRAU frame, starting from the parameter group in which the error was first detected. With reference to Figure 2, if, for instance, the first bit of octet 6 in a TRAU frame is detected to be zero, all parameters included in the frame are useless and buffer memories 43-47 are not updated at all by the parameters of this frame. Correspondingly, if it is not until in octet 34 of a TRAU frame that the first bit is detected to be zero, only subframe IV contains useless data; in this case, buffer memories 43-46 are updated normally, and the updating of buffer memory 47 is prevented. The buffer memories which are not updated contain the parameters of the preceding frame, which the decoder 42 then uses for decoding. As the parameters do not change, no undesired and disturbing sounds are produced to the speech signal as when useless parameters are used. In an embodiment of the invention, the para¬ meter XMAX which controls the level of speech is modified in the event of an error so that the level of the speech signal produced by the decoder 42 falls to zero level or by a predetermined attenuation co- efficient. This is achieved in such a manner that when the frame error signal 48 is active, the parameter in question is read to the decoder 42 from a different location in the memory than in normal operation.

Even in other respects the invention has been described above by way of example with reference to the GSM system. However, the invention is also suitable for use in other communications systems where speech encoding parameters are transmitted in data frames.

The figures and the associated description are intended merely to illustrate the present invention. In

their details the method and decoder according to the invention may be modified within the scope of the appended claims.