A description is given of an advanced 32-kb/s ADPCM (adaptive digital pulse code modulation) algorithm and its implementation in a single-chip custom LSI using 2- mu m CMOS technology. The performance of the proposed codec, in terms of voice and modems operating at 1200-9600 b/s, is also discussed. The results indicate that the codec provides speech quality equivalent to that of the CCITT G.721 ADPCM and is capable of transmitting V.23 1200 b/s and V.27 and V.29 4800 b/s signals over at least four asynchronous tandem links. It can also transmit V.29 modem signals at 7200 b/s over at least four and at 9600 b/s over two asynchronous tandem links, and V.32 signals at 9600 b/s over four asynchronous tandem links.>

This paper describes a performance evaluation of an advanced 32 kbit/s ADPCM hardware codec in terms of speech and high-speed voiceband data, V.27 and V.29 at 4.8, 7.2 and 9.6 kbit/s. The results indicated that the advanced 32 kbit/s ADPCM provided an equivalent subjective speech quality to that of the G.721 ADPCM recommended by the CCITT, and was able to maintain the transparency of these high-speed voiceband data at 7.2 kbit/s over at least four and at 9.6 kbit/s over at least two asynchronous tandem links, respectively. Furthermore, the advanced ADPCM can quickly recover a complete synchronous operation in the tandem connections, as burst errors or short-hits damage the digital links.

This paper presents an evaluation of the performance, with low-speed voiceband data at 300 bit/s, 1.2 kbit/s and 2.4 kbit/s, of a 16 kbit/s APC (Adaptive Predictive Coding) with maximum likelihood quantization(MLQ). By measuring modem bit error rates, the effect of a long-term predictor and influence of a subscriber line and channel bit errors over the APC-MLQ codec were investigated. The results indicate that the 16 kbit/s APC-MLQ can successfully transmit these low-speed voiceband data and is applicable to digital maritime satellite communication systems which must carry both speech and voiceband data up to 2.4 kbit/s. The performance has also been confirmed through field trials with G-III facsimile transmission at 2.4 kbit/s over a digital maritime satellite transmission system.

This paper presents a variable rate APC coding with Maximum Likelihood Quantization (MLQ) and noise-shaping filtering which covers a wide range of coding rates from 4.8 kbit/s to 16 kbit/s. At a rate lower than 8.0 kbit/s a new hybrid coding based on the ACP-MLQ and a new high-frequency regeneration are proposed. Simulation results indicate that the APC-MLQ having noise-shaping filters is very effective for achieving a high quality of speech in variable rate coding.

To cope with effects of channel bit errors, a robust quantizer design is proposed. A large leakage is variable initially inserted into the quantizer step size adaptation. In addition, two adaptive predictors in parallel are included in the feed- back loop of an ADPCM encoding system, in which the algorithm of the steepest descent method for predictor coefficient updating is modified to improve robustness. A 32-kbps 4-bit ADPCM encoder with a modified leaky adaptive quantizer and two 2nd order adaptive predictors operating in parallel was evaluated by calculating segmental and long-term averaged signal to distortion ratios. It is shown by computer simulations that this new ADPCM encoding scheme can result in more robust behavior than conventional ADPCM encoding scheme with a first order fixed predictor.

This paper describes a highly sensitive speech detector and a high-speed voiceband data classifier capable of discriminating between speech and voiceband data of a 4.8 kbit/s 8-phase PSK and 4.8 kbit/s 8-point QAM, and a 9.6 kbit/s 16-point QAM as described in a CCITT recommendation. The presence of a speech signal is detected by analyzing short-time energies, zero-crossing rates, and sign bit sequences of the input signal. The proposed speech detector, with a short hangover time of 32 ms, is able to reduce the average talk spurt activity in an international satellite link to 36 percent. This detector can also classify the detected speech into narrow-band or wide-band spectrum sounds or a low power sound for a variable rate ADPCM encoding. Discrimination between speech and high-speed voiceband data is based on short-time energies, a zeros-crossing rate and linear prediction coefficients of an adaptive predictor. Classification among a 4.8 kbit/s 8-phase PSK and 8-point QAM, and a 9.6 kbit/s 16-point QAM can also be performed by an average prediction gain and a coefficient of variation of the short-term amplitude distribution of the input signal. Discrimination of voiceband data was performed successfully, and erroneous discrimination of talk spurt of telephone speech as voiceband data were, respectively, four times for two two-party conversations lasting 5 minutes in an international satellite link. This is equivalent to less than 0.09 percent of the conversation time.

This paper describes a digital speech interpolationadaptive differential PCM bit reduction technique in which digital speech interpolation (DSI) is combined with ADPCM encoding. A highly sensitive speech detector, a voiceband data discriminator, and a variable rate ADPCM encoding are used to achieve a high compression ratio. The speech detector proposed in [1] detects speech signals above -51 dBm with 32 ms hangover time; average talk spurt activity of 36 percent was measured on fully loaded trunks in an international satellite link. Features of the speech power spectrum are used for adaptively controlling the bit length from 2 to 4 in an ADPCM speech encoder. Voiceband data are detected with 10 ms by the voiceband data discriminator. 5 bit ADPCM encoding is applied to voiceband data to maintain transparency through the DSI-ADPCM system. A DSI gain of 3 is expected as a result of the highly sensitive speech detection, the variable rate encoding technique, and the voiceband data discrimination. Speech and voiceband data are efficiently transmitted through an ADPCM encoding with either a 6 or 6.4 kHz sampling rate converted from an 8 kHz sampling rate. To avoid a band limitation as much as possible, a frequency shift manipulation on the voiceband channel is incorporated prior to the sampling conversion. Consequently, a total bit reduction gain of 7 to 4 is expected relative to a 64 kbit/s PCM transmission. Satisfactorily high quality of the processed speech has been obtained through computer simulations.

This paper describes DSI-ADPCM system in which a digital speech interpolation (DSI) system is combined with an adaptive differential PCM (ADPCM). The speech detector in DSI detects speech signals above -51 dBm with 32 ms hangover time, and attains 36 % reduction of average activity of a trunk in the international long distance lines. The adaptive bit assignment to the voiced and unvoiced sounds is adopted to cope with the freeze-out condition. These facts result in a DSI gain of 3. Furthermore, the 4 bit encoding is applied to DSI output at ADPCM with 6 kHz sampling attaining an ADPCM gain of 2.6. While the frequency of PCM speech signal with 8 kHz sampling is shifted 250 Hz down to lower range before the conversion of the sampling frequency from 8 kHz to 6 kHz is made to avoid quality degradation. Thus the total gain of 7 is expected when DSI and ADPCM are incorporated, and has already been demonstrated by computer simulation with satisfactorily high quality of speech.