Error detection/correction and fault detection/recovery – Pulse or data error handling – Digital data error correction
Utility Patent
1997-03-21
2001-01-02
Moise, Emmanuel L. (Department: 2784)
Error detection/correction and fault detection/recovery
Pulse or data error handling
Digital data error correction
C714S755000, C714S757000, C714S752000
Utility Patent
active
06170073
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to method and apparatus for digital communication. In particular, but not exclusively, to encoding and decoding for digital speech communication.
2. Background Art
In digital communication systems such as speech communication systems, the speech signal is typically processed in frames such that the speech signal is divided into about 20 ms time periods in the transmitter. For each of these periods, i.e., frames, a set of speech parameters is produced which are sent to the receiver. The parameter sets represent the speech signal for each time period. For simplicity, the sets of speech parameters produced for each frame are referred to here as speech parameter frames or speech frames.
In wireless speech communication systems such as in digital cellular systems, there are a number of channel conditions that severely affect the quality of transmission. It is therefore important to carry out error correction coding for the speech parameters. Error correction coding such as convolutional coding provide high error resilience during conditions which result in low numbers of errors but in conditions giving a high number of errors the error correction capability of any error correction coding will eventually be exceeded. Under conditions of severe transmission errors it is therefore important to detect the errors that have remained in the received speech parameters and to prevent these badly corrupted speech parameter frames from being used for synthesis of speech in the receiver. The speech parameter frames that are not used for speech synthesis due to their high level of errors are called bad frames. The received speech parameter frames that are used for speech synthesis are called good frames. If bad frames were used for synthesizing the speech signal in the receiver, highly annoying degradation's such as clicks and beeps would be produced in the synthesized speech.
In the receiver speech frames which will produce no effect or have a low effect on the speech quality must be substituted for the bad frames. The parameters for the substitute frames are produced by bad frame substitution techniques in which a repeated or extrapolated version of one or more of the previously received good speech frames is generated.
In prior-art coding systems such as the full rate (FR) and half rate (HR) coders of the Global System for Mobiles (GSM) system and in the full rate coder of the US Time Division Multiple Access (TDMA) system IS-54, an error detection code is used to detect errors remaining in the received speech parameters. In these systems a cyclic redundancy check (CRC) error detection code is used for some of the most important bits in the speech parameters. In the full rate speech channel of the GSM system, the 50 most important bits of a total of 260 speech parameter bits in each frame are covered by a 3-bit CRC. When a transmission error is detected among the 50 bits covered by the error detection code, the received speech frame is declared to be a bad frame. These frames are not used for speech synthesis but are substituted using well known bad frame substitution techniques.
Bad frame substitution can work well for one, two, even several consecutive lost speech frames. However, since speech sounds typically last only about 200-300 ms at the most, extrapolation of speech parameter frames can only be carried out successfully for about 10 to 15 lost frames. After this period, the decoder output should be silenced. Thus, in typical bad frame substitution such as that utilized for the full rate channel of GSM, given in GSM rec. 06.11, when several consecutive speech frames are substituted by extrapolated parameter values the synthesized speech signal is at the same time gradually attenuated and finally silenced. The attenuation is performed to prevent long artificial sounds being generated in the substitution process.
Current wireless systems do not produce speech which is of the same high quality of that produced in land line systems. A reason for the poor performance of prior art-communication systems in poor transmission conditions is that severe degradations in the synthesized speech occur due to a cumulative process from a multitude of errors in the less significant bits outside the error detection code.
In the current GSM system most of the speech parameter bits are left totally outside error detection. To extend the error detection code to cover the less significant bits also, so that any error in them is detected, would provide safe operation against transmission error but would make bad frame detection too sensitive for small errors. For example, a single bit error in the less important bits would cause the received speech parameter frame to be classified as a bad frame and to be lost. Bad frame substitution and speech muting would then occur too often in poor transmission conditions. There is the likelihood that speech synthesis would be completely cut off except for occasional error-free speech frames being passed to the speech decoder. These would not be able to synthesize any understandable speech but only to generate disturbing sounds.
To improve the quality of speech transmission various methods have been proposed. For example, besides using error detection codes, it is possible to obtain an estimate of the quality of the receiver speech parameter frames on the basis of reliability information available from the radio parts of a receiver. This information can be, e.g., signal to noise ratio measured for the radio channel or soft bit values obtained from the demodulator. Also error correction decoding, such as the Viterbi algorithm, may be used to produce some indication of how well error correction has succeeded. Due to low reliability, this kind of additional soft quality information can only supplement the error detection information obtained using real error detection codes and it is too vague to result in any considerable improvement in bad frame detection.
SUMMARY OF THE INVENTION
In accordance with a first aspect of the invention there is provided an encoder for digital communication, comprising; means for encoding digital signals representative of data, means for classifying the digital signals into first and second classes indicative of their influence on data quality, and means for error detection encoding capable of generating at least two error detection codes, wherein the at least two error detection codes respectively correspond to the first and second classes.
In accordance with a second aspect of the invention there is provided a decoder for digital communication, including: means for receiving encoded digital signals classified into first and second digital signal classes corresponding to importance of said digital signals, error detection decoding means for generating an error signal corresponding to respective digital signal classes, means for estimating the quality of the received digital signals responsive to the error signals generated by the error detecting means and means for determining the utility of the received digital signals dependent upon a result of the quality estimating means.
In accordance with a third aspect of the invention there is provided a method for digital communication encoding, comprising receiving digital signals representative of data, classifying the digital signals into first and second classes indicative of their influence on data quality, and generating at least two error detection codes respectively corresponding to the first and second classes.
In accordance with a fourth aspect of the invention, there is provided a method for digital communication decoding, comprising: receiving encoded digital signals classified into first and second digital signal classes corresponding to the important of said digital signals, decoding the received encoded digital signals for detecting an error corresponding to respective first and second digital signal classes, and estimating the quality of received digital signals based on a result of decoding the r
Haavisto Petri
Honkanen Tero
Jarvinen Kari
Vainio Janne
Lin Samuel
Moise Emmanuel L.
Nokia Mobile Phones (UK) Limited
Perman & Green LLP
LandOfFree
Method and apparatus for error detection in digital... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for error detection in digital..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for error detection in digital... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2449935